|
Citation |
- Permanent Link:
- http://ufdc.ufl.edu/AA00054895/00001
Material Information
- Title:
- A predictive model for academic success at the University of Florida
- Creator:
- McFaddin, Ronald Wheeler, 1934-
- Publisher:
- University of Florida
- Publication Date:
- 1971
- Language:
- English
- Physical Description:
- viii, 72 leaves. : ; 28 cm.
Subjects
- Subjects / Keywords:
- Academic achievement ( jstor )
Academic probation ( jstor ) College students ( jstor ) College transfer students ( jstor ) Colleges ( jstor ) High school students ( jstor ) Modeling ( jstor ) Predictive modeling ( jstor ) Transfer students ( jstor ) Universities ( jstor ) Counseling ( lcsh ) Dissertations, Academic -- Educational Administration and Supervision -- UF ( lcsh ) Educational Administration and Supervision thesis Ed. D ( lcsh ) Prediction of academic success ( lcsh ) City of Gainesville ( local )
Notes
- Thesis:
- Thesis (Ed. D.)--University of Florida, 1971.
- Bibliography:
- Bibliography: leaves 68-71.
- General Note:
- Manuscript copy.
- General Note:
- Vita.
Record Information
- Source Institution:
- University of Florida
- Holding Location:
- University of Florida
- Rights Management:
- The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
- Resource Identifier:
- 029529250 ( ALEPH )
14393879 ( OCLC )
Aggregation Information
- UFIR:
- Institutional Repository at the University of Florida (IR@UF)
- UFETD:
- University of Florida Theses & Dissertations
- IUF:
- University of Florida
|
Downloads |
This item has the following downloads:
|
Full Text |
A PREDICTIVE MODEL FOR ACADEMIC SUCCESS
AT THE UNIVERSITY OF FLORIDA
By
Ronald Wheeler McFaddin
A Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Education
UNIVERSITY OF FLORIDA 1971
Copyright by
Ronald Wheeler McFaddin
1971
ACKNOWLEDGMENTS
The writer wishes to acknowledge the assistance of the many persons who took an active interest in the preparation of this study.
The counsel and assistance of Dr. James L.
Wattenbarger, Chairman of the writer's Advisory Committee, are deeply appreciated. Sincere thanks are also extended to the members of the committee, Dr. Bert L. Sharp, Dr. Dayton Y. Roberts, Dr. Robert T. Frossard, and Dr. John H. James.
Appreciation is expressed to the many individuals of the Office-of Academic Affairs, especially Mr. Victor Yellen and Mr. Arthur Haller, the Office of Student Development, and the Board of University Examiners for their support and assistance in collecting data for this study.
The writer is also indebted to Dr. Charles M.
Bridges, Dr. Gerald R. Boardman, Mr. Gene A. Barlow, and Mr. William Graves for their advice and assistance during the study.
To his wife, Katharine, his son, David, and his
daughter,Laurie, the writer wishes to express his deepest gratitude for their love, devotion, patience, understanding, and assistance during the preparation of this study.
iii
TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS ..................................... iii
ABSTRACT ............................................ vi
Chapter
I. INTRODUCTION .................................. 1
The Problem .............................. 3
Assumption ............................... 6
Definition of Terms ........................ 6
Review of Related Literature ............... 8
Procedures ............................... 16
Summary .................................. 18
II. IDENTIFYING THE DATA ........................ 20
Introduction ... ......................... 20
Collection of Data ....................... 20
Computer Data Cards ...................... 23
Analysis of Data ......................... 24
Summary .................................. 27
III. DEVELOPMENT OF THE MODEL .................... 28
Introduction ...... ...................... 28
Developing the Model ..................... 28
Arts and Sciences ........................ 32
Business Administration .................. 36
Education ................................ 40
Journalism ................. ............. 44
Demonstration of the Predictive
Model .................................. 48
Summary .................................. 50
IV. IMPLICATIONS AND RECOMMENDATIONS ............ 52
Introduction ............................. 52
Implications ........................... 52
Weaknesses of the Model .................. 53
iv
TABLE OF CONTENTS-Continued
Chapter Page
IV. Use of the Predictive Model ............. 54
Further Research ........................ 55
Summary ................................. 56
Appendix
A. STUDENT REGULATIONS AT THE UNIVERSITY
OF FLORIDA .............................. 58
B. COMPUTER CARD DATA CODING .................. 66
BIBLIOGRAPHY ........................................ 68
BIOGRAPHICAL SKETCH ................................ 72
V
Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Education
A PREDICTIVE MODEL FOR ACADEMIC SUCCESS
AT THE UNIVERSITY OF FLORIDA By
Ronald Wheeler McFaddin
March, 1971
Chairman: Dr. James L. Wattenbarger Major Department: Educational Administration
The high attrition rate of students who enroll in
institutions of higher education has raised serious questions concerning the responsibility of higher educational institutions to increase the probability of academic success for these students. This study attempts to deal with one of these responsibilities, that of guidance, through the development of a predictive model for academic success.
Students enrolled in four colleges of the University of Florida as first-time juniors during the period Fall 1967 through Summer 1968 and who met the criterion of academic success or its antithesis were the sample for this study. The sample included 2,259 students. The term academic success was defined as the attainment of the bachelor's
V1
degree from the college entered at the beginning of the junior year, and its antithesis was the withdrawal of the student from the college while under academic probation or suspension. Academic records of the sample students were followed from their term of entry through Summer 1970. The data collected were taken from existing student records maintained by the University of Florida, and no new data were created for this study. The data were subjected to stepwise discriminant analysis by a computer technique of the BMD Biomedical Computer Programs package.
Four sets of models were produced (one set for each college) which may be used for the prediction of academic success or its antithesis. The program, on the basis of its developed model, predicted success or withdrawal for each of the students in the sample. Based on the actual performance of these students these predictions misclassified a number of cases. The percentages of misclassification ranged from 20 to 28 per cent for the four colleges.
The findings indicate that the probability of a student achieving academic success at the University of Florida might be directly related to his choice of college enrollment within the university. People who counsel students in their academic lives need to be aware of such possible different outcomes between the colleges within the institution.
vii
Those who counsel students should be cognizant
that a predictive model is a tool. Professional knowledge, judgment, and experience cannot be replaced by a predictive model for academic success, but can only be aided.
viii
CHAPTER I
INTRODUCTION
Of the students who entered higher educational
institutions over the past several decades, approximately 60 per cent eventually were graduated from some institution of higher education (17). Of the 40 per cent who did not graduate, how many could have been saved had administrators, counselors, or teachers known, through the use of some academic success predictor model, the probability of academic success for these students? Had these students been forewarned with individualized guidance, and thereby better able to prepare themselves, would the percentage of graduates have been much higher? It is the responsibility of higher educational institutions to become aware, on individual bases, of high risk students so that individual, special efforts can be made to reduce the probability of failure for these students to succeed academically.
Schroeder and Sledge studied factors related to collegiate academic success. In their review of recent studies, they found those studies in the minority which used specific collegiate averages and/or averages acquired beyond the first year of college. They also found that methodology was restricted primarily to simple and multiple
- 1 -
-2
correlations supplemented, in some cases,by tests of significance. They reported that intellective factors were found to be more predictive of collegiate achievement than non-intellective factors although the importance of the latter was not disputed (29).
Schroeder and Sledge reported that
multiple correlation coefficients, computed in an
effort to determine the best possible predictive combinations, did not exceed .71. Some studies, however, by
utilizing freshman grades as one of the predictors of achievement beyond the freshman year, were able to secure a multiple correlation beyond .80. (29, p. 98)
Such high correlations would be very valuable to a
counselor as predictors. The present study used both intellective and non-intellective factors as predictors of collegiate achievement. Freshman and sophomore grades were considered among the intellective factors in this study.
There is a need for a model at the University of Florida to predict the probability of academic success on the basis of common data available on student records. A counselor or other helping person would then be able to work individually with each potentially unsuccessful student to increase his probability of academic success. This study used a multivariate analysis computer program technique in the attempt to develop such a model to predict academic success at the University of Florida.
The Problem
Statement of the Problem
The purpose of this study was to develop a set of models for predicting academic success for students in selected colleges at the University of Florida. Justification for the Study
Although attempts have been made over the years to provide models for the prediction of academic success, only the recent advent of high-speed computers and the concomitant development of sophisticated techniques of data treatment have made this topic practicably researchable (6).
The problem for this study was a practical problem, as many administrators and counselors have been, and will be, faced by students with the question: What chance do I have? The development of a predictor model would provide a valuable tool for people who help and guide the academic lives of students. If facts, predicated by such a model, are presented to the student, it could be made quite clear to him just how much his own desire to succeed would really count. Often it is this desire, this motivation, on the part of the student which causes the difference between academic success and academic failure.
The United States needs as many of its students to
succeed academically as are able. The normal 40 per cent who
are lost represents a tragic waste which this nation cannot afford. If the 1967 Fall enrollment of 1,652,317 first-time freshmen in all public and private institutions in the United States were considered, the latest year reported by the OEO as reported in the Yearbook of Higher Education-1969, this would amount to nearly 661,000 youth lost (25, p. 382).
In research related to the junior college, it was
suggested by Thornton (34) and by Trent and Medsker (35) that for 35 per cent to 70 per cent of the students who are not academically successful,. their failure was to some extent due to a failure of the college. One of the major reasons given for failures was the lack of adequate guidance for students, and that generalization might be found to be true in senior institutions too. A predictor model for academic success could be a valuable tool in providing more adequate guidance.
At the present time, there is no tool being used at the University of Florida which considers the many variables included in this study.
Delimitations and Limitations
Delimitations:
1. This investigation was confined to students in the
junior and senior years in the colleges of Arts and Sciences,
Business Administration, Education, and Journalism at the University of Florida.
2. Only data made available to the researcher by the Office of Academic Affairs of the University of Florida were utilized in this study. The data available were limited to the period Fall 1967 through Summer 1970.
3. This investigation was limited to those variables indicated in the research design, namely: academic grade point average for freshman and sophomore years, Florida Twelfth Grade Test score or its equivalent, age of student in months, from birth to entry into his junior year, sex of student, native or transfer student, Florida or non-Florida resident, local residence, race of student, and marital status of student at the beginning of his junior year of study.
4. The analysis of data was confined to the stepwise
discriminant analyses of the investigated colleges by a computer technique of the BMD Biomedical Computer Programs package which produced a set of equations from which a set of predictor models of academic success could be evolved (12). Limitations:
This study was an ex post facto research study and
subject to the limitations of this method (20). It was also subject to the degree of the reliability of the records used to collect the data. Certain information which may be
-6
considered as predictors, e.g., a measurement of motivation,
was not available to the researcher for this study.
Assumption
It was assumed for the purpose of this study that
like grades are equal, e.g., the grades of "A" are equal
regardless of institutional, college, or professorial source.
Definition of Terms
The basic documents for the definition of terms were
The University Record, of the University of Florida, Undergraduate Catalog Issues for the years 1967 and 1970 (32, 33).
Academic probation-.That condition of formal recognition of the fact that a student is making unsatisfactory progress. For the purpose of this study, this term is inclusive of all degrees of academic probation as defined in the Undergraduate Catalogs (32, 33). For a complete explanation of this term see Appendix A.
Academic success-For the purpose of this study, this term will mean the attainment of the bachelor's degree from the college entered at the beginning of the junior year of study. This is a narrow definition of academic success, but the antithesis will also be narrow in that this group will include only those students who withdrew from their college while under academic probation or were suspended. The large group of students who might have withdrawn from the institution to attend another, or transferred to another college, or continued in attendance beyond the period of this study, or withdrew for any of a number of other reasons, were considered to be neither success nor failure; they were excluded from this study.
Academic withdrawal-For the purpose of this study, this term will be the antithesis of the term Academic Success.
Entry age-The age of the student in months, calculated from date of birth to date of entry in his junior year of study.
- 7
Florida resident-A person who is a citizen of the United States or a resident alien and who shall have resided and had his habitation, domicile, home and permanent abode in the State of Florida for at least twelve months immediately preceding his current registration (32, 33). Non-Florida resident includes all those students who do not meet the resident requirements as listed above. For a complete explanation of this term, see Appendix A.
Florida Twelfth Grade Test score-That score or its equivalent as converted from an acceptable general ability test, e.g., SAT or SCAT, which ranks the high school seniors of the State of Florida. The official name of this program is the Florida State-Wide Twelfth Grade Testing Program.
Grade point average-The average as determined by computing the ratio of grade points to credit hours attempted. Grade points are established by equating each credit hour as follows: A with 4, B with 3, C with 2, D with 1, and E, EW, I, WF, and X with 0.
InteZZllective factors-Those variables which are measured by ability, aptitude, or achievement tests, or some measure of prior achievement such as grade point average.
Junior year of study-For the purpose of this study, this term will include only those students who enter one of the included colleges for the first time and, by the University of Florida, are classified as 3AS, 3BA, 3ED, or 3JM.
Model-The models developed in this dissertation are represented by regression equations in which the weighted independent variables selected from the list above will predict the probability of the dependent variable, academic success. These equations were developed by using a BMD Program computer technique.
Native student-The student admitted to the Upper Division from the University College of the University of Florida who has taken not more than 10 credit hours from another institution. The limitation of 10 credit hours was also used by Walker in his study (36). This allows for the student who takes a course or two during the summer at another institution closer to home still to be considered a native student.
-8
Non-intellective factors-Those variables not included in the term intellective factors, e.g., entry age of student for junior year, sex of student, native or transfer student, marital status of student at beginning of junior year, Florida or non-Florida resident, on or off campus residence, and race of student.
Suspension and exclusion--For the purpose of this study, these terms may be used interchangeably. As stated in The University Record, of the University of Florida, the purpose of suspension from the University for academic reasons is to remove from the University those students who would not ultimately meet requirements for graduation if they continue at their current level of progress (33). For a complete explanation of these terms see Appendix A.
Transfer student-That student admitted to the Upper Division from any source other than the University College of the University of Florida, and those students admitted from the University College who transfer more than 10 credit hours from other institutions. This group includes students from public and private junior colleges, students from public and private senior institutions of Florida, and students from out-of-state senior institutions.
Upper division-The junior and senior years of the colleges included in this study.
Review of Related Literature
The development of a predictor model for academic
success was drawn from a review of such questions as: the
college attrition problem, methods of studying the attrition
problem, and types of analysis of those methods. A report
of the review of these questions is presented in this section.
The College Attrition Problem
An investigation by Cooper cited the results of a
governmental study conducted in 1938 by McNeely which dealt
- 9
with data collected from over fifteen thousand students in 25 universities. The attrition rate at the end of four years was found to be 62.1 per cent. He reported a number of students later enrolled in the same or another school and eventually received a degree, reducing the net attrition rate to 45.2 per cent (11).
A study reported by Iffert in 1958, with a sample of institutions, found that
slightly fewer than 40 per cent of students entering
the higher educational institutions in the study
graduated from the institution of original registration
in normal progression-that is, in a 4-year periodand perhaps 51 per cent graduated from some institution in this period. The conclusion that nearly 60 per cent of the students in the study eventually graduated from some institution of higher education
seems to be justified by the data. (16, p. 20)
Summerskill reported in 1962 that he reviewed 35 different studies that cited attrition rates for classes entering hundreds of varied colleges and universities from 1913 through 1962. He concluded that apparently the attrition rate had not changed appreciably during that 40-year period. While the rates varied from 12 per cent to 82 per cent in the 35 studies, about 40 per cent of college students, over the past several decades, have graduated on schedule and perhaps another 20 per cent graduated from some college at a later time (31).
Irvine reported that 49.5 per cent of entering freshmen at the University of Georgia in 1955 had graduated by
- 10
1963 from some institution. This was a period of eight years. It was found that women more frequently than men graduated "on time," but that in the long run, slightly more men obtained degrees from some college (18).
Walker studied transfer and native students in the upper division of the University of Florida. He reported that 62.5 per cent of native students and 32.2 per cent of transfer students who had entered the upper division in the fall term of 1966 had graduated by the end of the summer term of 1968 (36). The implication is that there may be sharp differences in attrition rates among different groups of a student body.
We may conjecture that the more restrictive admission policies since the mid-sixties have changed the attrition trend of the past several decades. However, there is no evidence in the literature to support such an idea. This nation cannot afford this continued waste of human talent, and it is fitting that efforts be made in the decade of the seventies which will improve the record by providing tools for predicting academic success and helping students make decisions appropriate to their needs. The high-speed computer will provide the ability for this undertaking to allow counselors and administrators a better understanding of individual students.
Methods of Studying the Attrition Problem
Atwell reported that early investigators of prediction usually examined the relationship of some intellective variable, e.g., an aptitude or achievement test score or some measure of prior achievement such as grade point average or rank-in-class, to a criterion or dependent variable which was nearly always some type of grade average. Later studies have included non-intellective factors in the research design and have often used a multi-statistical technique to measure the combined predictive value of two or more variables (2).
In addition to this type of study, other investigators have concentrated on non-intellective factors. Roudabush concluded that biographical information can contribute substantially to the prediction of academic criteria (28). The present investigation used non-intellective, biographical information as some of the variables studied.
Richards et at. used non-academic accomplishment such as participation in extra-curricular activities as predictors. These were found,however, to be largely independent of academic potential and achievement (26). Research using nonintellective variables also was stressed by Lavin in his book on predicting academic performance (21).
One of the more concise discussions of this subject was by Fishman in which he criticized research people for
- 12
simply adding non-intellective measures to the already existing test batteries. These additions were usually a personality test type instrument. Fishman argued that these instruments seemed to be "measuring something insufficiently dissimilar from whatever it is that our usual ,predictors are measuring" (13, p. 670). He explained his reasoning for taking this stand by a discussion of the usual predictive research method where
the most usual predictors are high school grades and scores on a standardized measure of scholastic aptitude. The usual criterion is the freshman average.
The average multiple correlation is approximately
.55. The gain in the multiple correlation upon adding a personality test score to one or both of the
usual predictors, holding the criterion constant,
is usually less than .05. (13, p. 699)
An example of a study which used non-intellective
variables, but of a discrete nature, rather than continuous, is that by Byron. He attempted to distinguish between successful and unsuccessful junior college students on the basis of such non-intellective variables as: residency (whether or not the student lived in the college district), age, sex, marital status, on or off campus housing occupancy, part time work, and others. His study found only sex and part time work to be significantly different for the two groups. He found that if the student were female and worked part time, the chances of being successful were greater. He recommended further investigation of non-intellective variables to increase the knowledge of student
- 13
behavior (7). The present study has included most of these same non-intellective variables. Types of Analysis
Hopper used multiple regression in his study and
the results generally supported the efficacy of considering grades when one wishes to predict grade success at the next level. But, more importantly, this study resulted in some rather substantial multiple R's, and introduced some non-intellective variables that had not heretofore been studied in any frequency (15). Those variables included study habits and attitude tests. His large multiple R's were done with Florida Twelfth Grade Test Scores.
A study by Rose and Elton concerned an investigation of personality differences between groups of students who withdrew from colleges at different times. They used multiple discriminant analysis which revealed significant differences among these groups of students. This was with respect to certain personality traits (27).
A study of demographic differences by Magoon and Maxwell, which made use of Chi-square test techniques, found significant differences within the colleges studied, but not between the colleges studied on the variables tested (22).
Bayes' review of the literature showed that the best prediction of subsequent academic performance is
- 14
obtained from multiple correlation in which a battery of intellective variables are used to predict the overall grade point average. He found that canonical correlation does not provide more adequate prediction of academic achievement, nor is it more logically defensible than multiple regression in the prediction of the academic achievement of college freshmen. His study also showed that personality variables do not significantly increase the percentage of predicted variance in academic prediction
(5).
It was shown by Maxwell that discriminant function and canonical variate analysis can be applied to a problem involving dichotomous variables (23). Woodward used both continuous variables and dichotomous data in his study for prediction of academic performance (37). The present study also used both continuous and dichotomous data in the search for a predictor model.
An instrument was produced by Ivanoff using discriminant analysis and continuous data for predicting freshman probationary students. Prediction tables were constructed for male and female students to serve as tools for personnel workers while helping students (18).
In a study by Conklin and Ogston, a selection of achievement, intellective, and personality tests were administered to 639 university freshmen in an attempt to
identify variables related to first year success. Correlation and regression analyses were conducted on the data. The results showed high school averages to be the best predictor while the other variables were shown to possess little predictive utility (9). The study by Gadzella and Bentall produced much the same results though with different groups (14).
Sims made his study in Florida and was able to
conclude that lower division grade point averages and scores from the Florida Twelfth Grade Testing Program were significant predictors of upper division grades (30). The study by Bauer et at. reached a similar conclusion in that they found the best single predictor of academic success to be past academic success (4). The present study included these same variables in the production of the predictor model.
A recent study by Clarke and Ammons considered
disadvantaged students. They found that selected intellective instruments are predictors of academic success for white males and females, but none of their cognitive measures proved to be significant for Negro males (8). The variable of race was considered in the present study.
Bashaw developed a "central prediction system" to forecast the success of junior college transfers in Florida universities. He considered intellective variables,
- 16
and no nonintellective variables in his study (3). The present study did consider intellective and nonintellective variables as well as transfer and native students as variables in developing a model for predicting academic success.
Justman reported that both precise and imprecise
measurements are used by counselor personnel. He cautioned that, because of the many controllable factors involved, predictions should be tentative. Predictions should be based on recent evidence which is relevant to the area of prediction, and should grow out of consideration of as many independent bases as are available. He added that, in the end, subjective evaluation will of course play a prominent part (19). However, it would be well to have a predictor tool as part of the evidence available. The development of such a predictor tool was the purpose of this present study.
Procedures
Study Design
This was an ex post facto research study in which several continuous and discrete independent variables obtained from records of students were observed. From these, the dependent variable of academic success was predicted.
- 17
The purpose of this design was to lead to the
development of sets of mathematical models built upon these independent variables. Such models could be useful as a tool in predicting academic success and guiding students' academic lives.
Sample
The sample for this study was all students who entered the colleges of Arts and Sciences, Business Administration, Education, and Journalism at the University of Florida in their junior year of study during the first academic year of the period of this study, and who met the criterion of Academic Success or its antithesis as defined above.
Data Collection
Data on each student included in this study were
supplied by the Office of Academic Affairs of the University of Florida. The data included most of the variables considered by previous researchers to be the "best predictors" of academic success for college students.
The data included in this study were as follows:
1. Lower division grade point average (freshman and
sophomore years)
2. Florida Twelfth Grade Test score or its equivalent
3. Age of student at beginning of junior year
- 18
4. Sex of student
S. Native or transfer student
6. Florida or non-Florida resident
7. Local residence 8. Race of student
9. Marital status of student at beginning of the
junior year
Data Treatment
The data collected were subjected to the stepwise discriminant analyses by a computer technique of the BMD Biomedical Computer Programs package. The College of Education Research Service Center has adapted these procedures to the University's IBM 360 computer for application in educational research.
The data for each college were given separate
treatment. The result of the use of the treatment technique was the development of a prediction model for academic success for each college included in this study. The models show a weight factor for each of the variables included, ranking these from most significant to least significant.
Summary
From the review of related literature it appears that almost half of the students who enter institutions
- 19
of higher education fail to complete their studies and receive a baccalaureate degree. Additional information, such as a predictor model for academic success, may help more students to succeed.
Previous research emphasis has been based on the
assumption that academic success predicts academic success. While this study considers this point, it goes beyond and attempts to consider other variables that may be indicative. Recent concern for disadvantaged students and others who have not achieved academic success has raised the question of relying solely on this assumption and pointed out that we require better answers than have been previously found if we are to properly do our jobs as educators.
Chapter I is an introduction to this study which outlines the problem and includes a review of the related literature. Chaper II describes the procedure used to develop the model for the prediction. Chapter III is an analysis of the data resulting from the study and an outline of the development of the model. Chapter IV introduces some theoretical considerations and implications and suggests use of the predictor model for academic success.
CHAPTER II
IDENTIFYING THE DATA
Introduction
This was an ex post facto research study in which several continuous and discrete independent variables obtained from records of students were observed. From these independent variables, the dependent variable of academic success was predicted. The purpose of this study was to develop a set of mathematical models (the predictive model) for predicting academic success for students in selected colleges at the University of Florida.
Collection of Data
Data collection for this study was made possible by the Office of Academic Affairs of the University of Florida. The design of the study required examination of complete academic records of all the students who entered each of the four colleges included in the study as juniors during the period of Fall 1967 through Summer 1968. These colleges were Arts and Sciences, Business Administration, Education, and Journalism.
The data utilized in this study were taken from
existing, individual student records as maintained by the
- 20 -
- 21
Registrar and Admissions Office of the University of Florida. All of the information headings maintained on these records were studied by the researcher and the selection of those included in the study was based on data used by previous researchers as cited in the review of the literature, and the commonality of availability. It is important to note that no new data were created for this study. All data used were already available on student records maintained by the University of Florida.
Often many of the records of transfer students do not include as much information as the records of native students. It is important that any data used in the development of a predictor model be commonly available on student records. The inclusion of variables which are commonly found to be missing would limit the usefulness of any success predictor model. In addition, certain kinds of information that might be useful in a predictor model, such as the record of part time employment by a student, simply are not maintained on these records.
The population for this study was compiled from the cumulative tapes of student records of the Registrar and Admissions Office. The computer print-out of this compiled population included the academic record of each individual student from his time of entry at the beginning of this study through the term of Summer 1970. It was
- 22
possible to observe individual progress by academic term, including end of term actions taken regarding each student, for a maximum of twelve academic terms. As these academic terms were academic quarters, normal progress would have been a total of six academic terms to graduation, that is, the completion of the junior and senior years of college. This population included 3,156 students.
Inspection of this population led to the selection of the study sample. Inspection was made on the basis of the criterion Academic Success or its antithesis, Academic Withdrawal, as defined above. This produced a sample of 2,344 students who had met these qualifications. A print-out of the complete record of each of these students was extracted by computer from the various cumulative tapes maintained by the Office of Academic Affairs and the Registrar and Admissions Office. These data showed specific information to be missing which were necessary to this study. The largest amount of missing data was in the Florida Twelfth Grade Test score section. Also high in the missing area was the section for grade point average for freshman and sophomore years. It should be noted that these data were only missing from the cumulative tapes and not the individual student records.
Because of the missing data, an individual records search was necessary for approximately 1,000 students.
- 23
This examination of individual student records was made possible by the Office of Student Development which allowed the researcher access to their files. Only eighty-five students who were included in the original sample could not ultimately be included in the study because of the lack of information on their individual records. This left a final sample of 2,259 students for the study.
Over 200 Florida Twelfth Grade Test score
equivalents were provided by the Board of University Examiners for ACT, SAT, and SCAT scores for individual students. The Florida Twelfth Grade Test was created by the Board of University Examiners and comparative studies by them over the years have enabled them to make up equivalent scores for these other examinations which are statistically reliable. This conversion was necessary for standardization so that these individuals could be included in the study as defined above.
Computer Data Cards
Data cards were keypunched for each student and included the following information for this study:
1. Lower division grade point average (freshman and
sophomore years)
2. Florida Twelfth Grade Test score or equivalent
3. Age of student in months, from birth to entry into
junior year (possible starting dates for the junior
year were: 9/67, 1/68, 4/68, or 6/68)
- 24
4. Sex of student
5. Native or transfer student (native students have
less than 11 transfer hours of credit)
6. Resident (Florida, non-Florida, or alien)
7. Residence (campus, fraternity or sorority, campusmarried housing, or off campus)
8. Race of student (white or Negro)
9. Marital status of student at beginning of the junior
year of study (single, married, divorced, or widowed)
Additional information was keypunched, some of which was needed for identification purposes during the analysis of the data, such as:
1. College membership (Arts and Sciences, Business
Administration, Education, or Journalism)
2. Academic success or its antithesis, academic
withdrawal
3. Social security number
See Appendix B for the coding of the data cards.
Analysis of Data
The method of analysis of these data was a
multivariate analysis of the BMD Biomedical Computer Programs package developed at the University of California
(12). The particular program used was identified as BMDO7M Stepwise Discriminant Analysis. This program was designed for use where data are categorized by case and the cases are designated as belonging to one of two or more groups. Linear sums of the variables are determined
- 25
which classify the cases into groups. This program proceeds in a stepwise manner by forming linear sums of first one, then two, three, etc., variables. At each step the variable added is the one which gives the greatest improvement in classification. The variable entered is selected by the first of the following equivalent criteria:
1. The variable with the largest F value. (This
statistic is the Fisher F. The use of the F
value in this procedure gives the largest likelihood of difference with all the variables in the equation in order to maximize the difference in
the groups.)
2. The variable which when partialed on the
previously entered variables has the highest
multiple correlation with the groups.
3. The variable which gives the greatest decrease
in the ratio of within to the total generalized
variances.
For a more complete description of this program see Dixon (12, p. 214a 214b).
The problem in discriminant analysis is getting a linear combination of variables which distinguish, better than any other linear combination, between the two or more chosen groups. Descriptions of discriminant function analysis can be found in Anderson (1), and Cooley and Lohnes (10). Discriminant analysis is very similar in procedure to multiple regression analysis. The principal difference between the two techniques lies in the nature of the dependent or criterion variable. In regression analysis, the dependent variable is a continuous variable taking
- 26
infinitely many values; e.g., grade point average. In discriminant analysis, the dependent or criterion variable is group membership, which may or may not be on a continuum (it may be discrete). Thus discriminant analysis begins with defined groups and seeks to weight scores in order to achieve a maximal difference among group means; e.g., with the previously defined groups Academic Success and Academic Withdrawal.
The data were separated by college so that four
sets of models might be produced, one for each college in the study. Each of these four groups was further broken into two groups; the group who experienced academic success, and the group who met the antithesis of academic success (academic withdrawal). These groups were composed of the following individuals:
College Success Withdrawal Total Arts and Sciences 752 149 901 Business
Administration 387 88 475 Education 570 39 609 Journalism 247 27 274
Total 1,956 303 2,259
The BMD Biomedical Computer program was able to
perform the stepwise discriminant analysis for each of the colleges in one operation. Therefore, all data were loaded,
- 27
together with the appropriate control cards, and submitted to the University of Florida's IBM 360 computer for analysis by the above described program.
Summary
The researcher used records of 2,259 students for
the study sample. The nine variables included in the study were taken from the data bank maintained by the University of Florida and no new data were created for this study. Therefore, the study could be repeated for any group of students an investigator might select. The analysis tool, the BMDO7M Stepwise Discriminant Analysis program, is an established, economical, analysis program that is widely available to investigators.
CHAPTER III
DEVELOPMENT OF THE MODEL
Introduction
The data collected for this study were submitted, in four parts, to the University of Florida's IBM 360 computer and subjected to discriminant function analysis using the BMDO7M computer program. The colleges included in the study, namely Arts and Sciences, Business Administration, Education, and Journalism, made up these four parts. This portion of the report will consider the results of the analysis described above and examine each college individually together with the sets of mathematical models (the predictive model for academic success) that evolved from this analysis.
Developing the Model
The sample data for the four colleges included
the records of 901 students in Arts and Sciences, 475 students in Business Administration, 609 students in Education, and 274 students in Journalism.
The analysis of the data by the program led to the construction of a set of mathematical models for
- 28 -
- 29
each college -one set of models for the prediction of academic success, and another for the prediction of academic withdrawal.
The models consist of a constant (K) and a function for each of the variables (X) included in the study which when added together result in the prediction. The predictive model for each college consists of a set of two models, one for predicting success (S) and the other for predicting withdrawal (W).
The models permit one to calculate figures for
both academic success and its antithesis, academic withdrawal, for any student for any of these colleges. This is possible because once weights have been established for the predictor variables by means of discriminant analysis the application of the weights to assign future cases to the groups tends to reproduce the composition of the original groups with respect to the measured variables. These calculated figures are the values for the discriminant function for the tested student and the smaller of the two figures is chosen as the predicted result. That is, whichever is smaller, the (S) or the (W), is the prediction of success or withdrawal for that individual student.
- 30
Discriminant analysis seeks to make the predefined groups maximally different and to facilitate the assignment of new subjects to one of the given groups. If one merely wished the groups to have different means, one could find the means and multiply them by constants of whatever size one wished. To be meaningful, the difference in group means should be maximal relative to the score spread in each group. This is the reason for the program establishing the specific constants (K) as shown in these mathematical models for each.college.
If chance alone were applied to the data of this study, 100 per cent of the variability would be accounted for by chance. If one uses these models, 27 to 42 per cent of the variability is no longer accounted for by chance, but can be attributed to the variables included in this study.
When one considers the overall college memberships in the data sample it will be noted that the models developed by the program correctly predicted academic success or academic withdrawal 72 to 79 per cent of the time for the individual colleges. This prediction was done on the 2,259 students included in the sample. The discrimination might be better described as the percentage of cases which are misclassified by the program of
- 31
the total sample. These percentages ranged from 20 to 28 per cent for the four colleges.
The efficiency of discrimination may be best
expressed in correlational form. An appropriate question for this test might be: "What is the relationship (correlation) of the actual group membership (the groups as selected on the basis of the criterion variable) to the ability of the computer developed model to predict?" An appropriate correlation technique for this test is the use of phi(c). Phi is a derivation of the Pearson r for dichotomous data and the data in this study may be considered dichotomous for statistical purposes. For a
2 x 2 table, a possible display of the results for each college of the predictive result (2) compared with (x) the actual result (2), 4 is directly related to the chi-square (X2) statistic. For a discription of these techniques see Ostle (24), and Wyatt and Bridges (38).
As a preliminary step in the development of phi, a chi-square coefficient of correlation was performed on the results of the analysis. This statistic was found to be significant at the .005 level of confidence for each of the colleges in the study. This indicates a high degree of relationship between the predicted results based on the above predictive model for academic
- 32
success and the observed results based on the criterion variable.
In order to determine the amount of variance
accounted for with this predictive model it was necessary to calculate phi (maximum) for each of the colleges. This was done so that the ratio of phi to phi(max) (4/max) might be calculated. In order to assess the percentage of variance accounted for one must square the / max ratio. This value indicates that this percentage of the variance was accounted for by the variables included in the predictive model. These values ranged from 27 to 42 per cent in this study. Any correlation coefficient when squared may be interpreted as a percentage of variance accounted for. For a discussion of variance see Kerlinger
(20), Ostle (24), or Wyatt and Bridges (38).
Arts and Sciences
The sample data for the College of Arts and Sciences included 901 students. By the criterion of academic success, described above, 752 students experienced success and 149 students experienced withdrawal, or the antithesis of academic success.
The analysis of the data collected on each of
these students led to the construction of a set of mathematical models:
- 33
S = -115.58421(K) + 0.09268(X1) + 0.11371(X2)
+ 0.15270(X3) + 5.37714(X4) + 8.58541(X5) + 9.63046(X6) + 1.45148(X7) + 55.44241(X8)
- 0.50514(X9)
W = -111.83672(K) + 0.07800(X1) + 0.10488(X2)
+ 0.16029(X3) + 6.31623(X4) + 9.51001(Xs) + 9.75150(X6) + 1.00025(X7) + 54.22322(X8)
- 0.16460(X9)
where:
S = Academic success
W = Academic withdrawal
K = A constant
X, = Lower division grade point average
X2 = Florida Twelfth Grade Test score or equivalent
X3 = Age of student at beginning of junior year
X4 = Sex of student
Xs = Native or transfer student
X6 = Florida or non-Florida resident
X7 = Local residence Xe = Race of student
X9 = Marital status of student at beginning of junior
year
The model for success for this college predicted correctly 77.5 per cent of the time, and the model for withdrawal predicted correctly 69.5 per cent of the time.
- 34
When one considers the overall college membership in the sample this set of models predicted correctly 76.2 per cent of the time.
The stepwise discriminant analysis loaded each
variable for this college in descending order of whichever one would give the greatest improvement in classification. That is, the most significant variables for classification were loaded first. These variables were loaded in the following order:
X2, Xi, X3, X7, X4, Xs, X8, X9, X6
After all the variables had been entered, the
cumulative approximate F ratio was found to be 26.51431 with 9 and 891 degrees of freedom. This is statistically a significant F beyond the .01 level, indicating little chance of these variables not being significant in the predictive model.
A chi-square coefficient of correlation test was performed on the results of the analysis as a preliminary step in the development of phi. This was found to be: X2 = 131.8840, which was found to be significant at the .005 level of confidence. The phi coefficient was found to be: 4 = 0.3825. This indicates a high degree of relationship between the predicted results based on the above
- 35
predictive model for academic success and the observed results based on the criterion variable.
In order to determine the percentage of variance accounted for by this predictive model, phi (maximum) was calculated. This was found to be: 4max = 0.6749, and the ratio of phi to phimax was calculated to be: 4/4max = 0.5668. This correlation coefficient when squared may be interpreted as a percentage of variance accounted for. This value indicates that 32.12 per cent of the variance was accounted for by this predictive model for academic success.
Student Profile
An analysis of the means of the variables included permits one to generalize the profile of these groups of students in the College of Arts and Sciences, those who succeed academically, and those who withdraw. Those students who succeeded had a higher lower division grade point average than those who withdraw. They also had a higher Twelfth Grade Test score than those who withdrew. Age was negative in its influence as the group who withdrew was older than the group who succeeded. The female student also withdrew from this college more often than did the male, as did the transfer student more than the native student. There was virtually no difference in success and
36
withdrawal on the variable of residence or non-residence of the state of Florida. There was a difference in the two groups on the variable of local residence, indicating more successful students live in on campus married housing or off campus than in campus single housing or in a fraternity or soroity. The variable of race was virtually meaningless for this study as almost all students included in the study were white. Actually, more students refused to disclose their race on their university records than there were Negroes in the study. The variable of material status seems to indicate that the single student who had never married had the best chance for success in this college.
Business Administration
The sample data for the College of Business
Administration included 475 students. Of this sample, 387 students met the criterion of academic success, described above, and 88 students met the criterion of academic withdrawal, the antithesis of academic success.
The analysis of the data for this college by
the BMDO7M program led to the construction of a set of mathematical models for the prediction of academic success or academic withdrawal for this college:
-37
S = -169.13904(K) + 0.17927(X1) + 0.08873(X2)
+ 0.24582(X3) + 36.05937(X4) + 4.49521(Xs) + 5.22244(X6) + 3.02595(X7) + 86.35260(X8)
+ 0.46761(X9)
W = -160.19824(K) + 0.16564(X1) + 0.08064(X2)
+ 0.24600(X3) + 35.90738(X4) + 5.50448(Xs) + 5.41641(X6) + 2.40330(X7) + 84.93285(X8)
- 0.79382(X9)
where:
S = Academic success
W = Academic withdrawal
K = A constant
X, = Lower division grade point average
X2 = Florida Twelfth Grade Test score or equivalent
X3 = Age of student at beginning of junior year
X4 = Sex of student
Xs = Native or transfer student
X6 = Florida or non-Florida resident
X7 = Local residence X8 = Race of student
X9 = Marital status of student at beginning of junior
year
The model for success for this college predicted correctly 74.2 per cent of the time, and the model for withdrawal predicted correctly 68.2 per cent of the time.
- 38
When the overall college membership in this sample is considered, this set of models predicted correctly 73.1 per cent of the time.
The stepwise discriminant analysis loaded each
variable for this college in descending order of whichever one would give the greatest improvement in classification. That is, the most significant variables for classification were loaded first. These variables were loaded in the following order:
X2, X7, XI, Xs, Xg, Xe, X6, X4, X3
After all the variables had been entered, the
cumulative approximate F ratio was found to be 11.83648 with 9 and 465 degrees of freedom. This is statistically a significant F beyond the .01 level, indicating little chance of these variables not being significant in the predictive model.
A chi-square coefficient of correlation test was performed on the results of the analysis as a preliminary step in the development of phi. This was found to be: X2 = 57.544, which was found to be significant at the .005 level of confidence. The phi coefficient was found to be 4 = 0.3480. This indicated a high degree of relationship between the predicted results based on the above
- 39
predictive model for academic success and the observed results based on the criterion variable.
In order to determine the per cent of variance
accounted for by this predictive model, phi (maximum) was calculated. This was found to be: 4max = 0.66017, and the ratio of phi to phimax was calculated to be: 4/4max = 0.52004. This correlation coefficient when squared may be interpreted as a percentage of variance accounted for. This value indicates that 27.04 per cent of the variance was accounted for by this predictive model for academic success.
Student Profile
An analysis of the means of the variables included in the study permits one to generalize the profile of these two groups of students in the College of Business Administration, those who succeed academically, and those who withdraw. Those students who succeeded had a higher lower division grade point average than those who withdrew. They also had a higher Twelfth Grade Test score than the group who withdrew. The group who withdrew was older than the group who succeeded, so age appears to be a negative influence for success in this college. The male student succeeded more often than did the female in this college, but there was very little difference in the means
- 40
of these two groups. The native student succeeded more often than did the transfer student. There was virtually no difference in the means of the two groups with respect to the Florida or non-Florida resident variable, although more who withdrew were non-Florida residents. With respect to local residence, the means indicated that the student who succeeded lived in campus married housing or off campus, rather than in campus single housing or in a fraternity or sorority. The variable of race was virtually meaningless for this college in that almost all students included in the sample were white. The variable of marital status seems to indicate that the single student who had never married was more likely to withdraw than the married, or widowed, or divorced student.
Education
The sample data for the College of Education included 609 students. By the criterion of academic success, described above, 570 students experienced success and 39 students experienced withdrawal, or the antithesis of academic success.
The analysis of the data collected on each of
these students by the BMDO7M program led to the construction of a set of mathematical models for the prediction of academic success or academic withdrawal for this college:
- 41
S = -106.61400(K) + 0.10421(Xl) + 0.06344(X2)
+ 0.12112(X3) + 8.35981(X4) + 8.71175(Xs) + 17.29904(X6) + 0.91885(X7) + 42.93976(X8)
- 5.84442(X9)
W = 99.00226(K) + 0.09207(X1) + 0.05771(X2)
+ 0.11396(X3) + 9.17078(X4) + 9.10707(Xs) + 16.98534(X6) + 0.53092(X7) + 41.84727(X8)
- 5.02412(X9)
where:
S = Academic success
W = Academic withdrawal
K = A constant
X, = Lower division grade point average
X2 = Florida Twelfth Grade Test score or equivalent
X3 = Age of student at beginning of junior year
X4 = Sex of student
Xs = Native or transfer student
X6 = Florida or non-Florida resident
X7 = Local residence X8 = Race of student
X9 = Marital status of student at beginning of junior
year
The model for success for this college predicted correctly 71.9 per cent of the time, and the model for withdrawal predicted correctly 71.9 per cent of the time.
- 42
When the overall college membership in this sample is considered, this set of models predicted correctly 71.9 per cent of the time.
The stepwise discriminant analysis loaded each
variable for this college in descending order of whichever one would give the greatest improvement in classification. That is, the most significant variables for classification were loaded first. These variables were loaded in the following order:
X2, X7, XI, X4, X8, Xg, X3, Xs, X6
After all the variables had been entered, the cumulative approximate F ratio was found to be 4.39773 with 9 and 599 degrees of freedom. This is statistically a significant F beyond the .01 level, indicating little chance of these variables not being significant in the predictive model.
A chi-square coefficient of correlation test was performed on the results of the analysis as a preliminary step in the development of phi. This was found to be: X2 = 32.7017, which was found to be significant at the .005 level of confidence. The phi coefficient was found to be: 4 = 0.2317. This indicated a high degree of relationship between the predicted results based on the above
-_43
predictive model for academic success and the observed results based on the criterion variable.
In order to determine the per cent of variance
accounted for by this predictive model, phi (maximum) was calculated. This was found to be: 4max = 0.39140, and the ratio of phi to phimax was calculated to be: p/max = 0.59197. This correlation coefficient when squared may be interpreted as a percentage of variance accounted for. This value indicates that 35.04 per cent of the variance was accounted for by this predictive model for academic success.
Student Profile
An analysis of the means of the variables
included in the sample of the students of the College of Education permits one to generalize a profile of these two groups of students, those who succeed academically, and those who withdraw. Those students who succeeded had a higher lower division grade point average and higher Florida Twelfth Grade Test score than those students who withdrew. They were also older than the group who withdrew. The female student withdrew more often than did the male student. The native student succeeded more often than did the transfer student, as did the Florida resident more than the non-Florida resident. The means
- 44
of the variable of local residence indicate that the successful student lives in campus married housing or off campus, rather than in campus single student housing or in a fraternity or sorority. The variable of race was meaningless for this college due to the small number of Negro students in the sample. The variable of marital status seems to indicate that the single student who had never married had the best chance for success in this college, but there was very little difference for the other classifications of this variable.
Journalism
The sample data for the College of Journalism
included 274 students. Of this college sample, 247 students met the criterion of academic success, described above, and 27 students met the criterion of academic withdrawal, the antithesis of academic success.
The analysis of the data for this college by the BMDO7M program led to the construction of a set of mathematical models for the prediction of academic success or academic withdrawal for students of the College of Journalism:
- 45
S = -125.20964(K) + 0.22712(X1) + 0.09616(X2)
+ 0.14281(X3) + 3.88128(X4) + 7.29087(Xs) + 7.99617(X6) + 3.99100(X7) + 55.19489(X8)
+ 6.86018(Xg)
W = -121.91797(K) 0.20694(X1) 0.08989(X2)
- 0.15107(Xs) 4.84507(X4) 9.16011(Xs) 8.00885(X6) 3.36210(X7) 54.47794(X8)
- 5.522ss55(X9)
where:
S = Academic success
W = Academic withdrawal
K = A constant
X, = Lower division grade point average
X2 = Florida Twelfth Grade Test score or equivalent
Xs = Age of student at beginning of junior year
X4 = Sex of student
Xs = Native or transfer student
X6 = Florida or non-Florida resident
X7 = Local residence X8 = Race of student
X9 = Marital status of student at beginning of junior
year
The model for success for this college predicted correctly 79.8 per cent of the time, and the model for withdrawal predicted correctly 74.1 per cent of the time.
- 46
When the overall college membership in this sample is considered, this set of models predicted correctly 79.2 per cent of the time.
The stepwise discriminant analysis loaded each
variable for this college in descending order of whichever one would give the greatest improvement in classification. That is, the most significant variables for classification were loaded first. These variables were loaded in the following order:
Xs, XI, X7, X3-, X2, X9, X4, X8, X6
After all the variables had been entered, the cumulative approximate F ratio was found to be 6.41046 with 9 and 264 degrees of freedom. This is statistically a significant F beyond the .01 level, indicating little chance of these variables not being significant in the predictive model.
A chi-square coefficient of correlation test was performed on the results of the analysis as a preliminary step in the development of phi. This was found to be: 2 =37.0809, which was found to be significant at the .005 level of confidence. The phi coefficient was found to be: 4 = 0.36787. This indicates a high degree of relationship between the predicted results based on the
- 47
above predictive model for academic success and the observed results based on the criterion variable.
In order to determine the percentage of variance accounted for by this predictive model, phi (maximum) was calculated. This was found to be: 4max 0.56435, and the ratio of phi to phimax was calculated to be: 4/5max
- .65190. This correlation coefficient when squared may be interpreted as a percentage of variance accounted for. This value indicates that 42.50 per cent of the variance was accounted for by this predictive model for academic success.
Student Profile
An analysis of the means of the variables
included in the sample for the College of Journalism permits one to generalize the profile of these two groups of students in this college, the group who succeed academically and the group who withdraw. Those who succeeded had a higher lower division grade point average, higher Florida Twelfth Grade Test score, and were younger than the group who withdrew. The male student was more successful than the female student. The native student was more successful than the transfer student, and the nonFlorida resident student was more successful than the Florida resident student. More successful students in
- 48
this college lived in campus married housing or off campus than lived in on campus single housing or in a fraternity or sorority. Race was a meaningless variable due to the few Negro students enrolled. The variable of marital status seems to indicate the successful student is married.
Demonstration of the Predictive Model
For a demonstration of how the predictive model for academic success wil work, three student data cards (identified as A, B and.C) were selected from the data deck and processed for each of the four colleges. The students and their coded variables are as follows:
Student A Student B Student C X, 219 210 336 2 310 376 340 Xs 241 259 308 X4 2 2 2 Xs 1 2 1 X6 4 4 4 X7 4 1 4 Xe 1 1 1 X i 1 2
- 49
See the sets of mathematical models above and
Appendix V (for data coding) for complete identification of the coded variables and the procedure for arriving at the predictions.
For student A the sets of mathematical models
indicate that one should have expected little chance for academic success in either of the Colleges of Arts and Sciences or Journalism. The predictive model also indicates that this student would have been unlikely to achieve academic success in the College of Business Administration, but the prediction (indicated by the size of the difference between the discriminant functions) is not so strong as for the other two colleges. The model does predict that this student would have been likely to achieve academic success in the College of Education. The observed result for this student was that she did not achieve academic success as a student in the College of Arts and Sciences.
For student B the predictive model indicates
that one could have expected the achievement of academic success in any of the four colleges included in the study. It does indicate that the chance for success in the College of Journalism should have been expected to be less than in any of the other three colleges. The observed
- 50
result for this student was that she did not achieve academic success as a student in the College of Journalism.
For student C the calculations from the predictive model indicate that one could have expected this student to have not succeeded academically in any of the four colleges in the study. It does indicate that the most likely chance for academic success would have been in either the College of Arts and Sciences or the College of Education as the prediction for academic withdrawal for these two colleges was not strong. The observed result for this student was that she did achieve academic success as a student in the College of Arts and Sciences.
Summary
The researcher presented an analysis of the data resulting from the study and the development of the predictive model for academic success in this portion of the report. Each college was treated separately with an analysis of the data and development of specific sets of mathematical models.
When one considers the overall college
memberships in the data sample it will be noted that the model correctly predicted academic success or academic withdrawal 72 to 79 per cent of the time for the individual colleges. The percentage of misclassified cases ranged from 20 to 28 per cent for the four colleges.
In order to determine the amount of variance
accounted for with this predictive model a derivation of the Pearson r for dichotomous data (phi) was calculated for each college. This value indicates the percentage of variance accounted for ranged from 27 to 42 per cent for the four colleges.
In three of the four colleges studied, Arts and Sciences, Business Administration, and Education, the Florida Twelfth Grade Test score had the heaviest loading. In the fourth college, Journalism, the native or transfer variable received the heaviest loading. The resident of Florida variable was generally the lowest factor in the loading.
Generally speaking, the model will allow one to predict academic success, based on these variables, to a sufficient degree that they could be useful to those who counsel students in their academic lives. The possibility of failure by the model was demonstrated by the prediction made on randomly selected student C. The demonstration also showed that there are differences in colleges within the university and the selection of a college by a student has a significant influence on the probability of achieving academic success by a student.
CHAPTER IV
IMPLICATIONS AND RECOMMENDATIONS
Introduction
This section includes a discussion of theoretical implications developed from the study. It considers weaknesses of the predictive model for academic success and makes recommendations for its use. Recommendations are also made for further study.
Implications
The strongest point to be made from this study is
that the probability of a student achieving academic success at the University of Florida may be directly related to his choice of college enrollment within the university. Previous research seems to have been concerned with the overall population of a university and has not looked at the individual colleges within the overall structure. The sets of mathematical models for the four colleges included in this study point out that differences exist between colleges within the university. New colleges have used different admission requirements. The demonstration of the model shown above makes this point quite clear with Student A. It would
- 52 -
- 53
appear that this student might have achieved academic success had she been counseled to enter the college suggested by the predictive model.
People who counsel students in their academic lives need to be aware of such possible different outcomes within the institution. The best possible choice, helped to be made by the use of the model, might make the difference between academic success and withdrawal from the institution.
Weaknesses of the Model
The model, statistically, has a built-in weakness
due to the use of one of the variables. The Florida Twelfth Grade Test score is the sum of five percentile scores attained by the student on each of five separate academic area tests. The justification for including this variable is that it is commonly used throughout the state of Florida and particularly at the University of Florida. It is used to rank students and thereby serves as a major criterion upon which to base admission to the University. This is especially true at the freshman level where a score of 300 is considered to be the minimum desired score. The statistical use of such a score (the sum of percentile scores) is open to question. The only defense offered is that this is common, acceptable practice in the state of Florida for this test series. It
- 54
would have been better to use raw scores, or a standard score in their place.
Previous researchers have found the variable of
race to be a significant discriminator. This was cited in the review of related literature. There were too few Negroes in the sample to allow this variable to have meaning in this study. The reason for this is that apparently few Negroes were enrolled as students at the University of Florida during the time of the study. This variable may become much more significant as a predictor in the years ahead.
Use of the Predictive Model
The predictive model for academic success should be used at the University of Florida by those persons who counsel students in the selection of a college for study. It should be made clear to students that differences do exist between colleges with respect to the individual student's probability for academic success.
The model should also be used within the four
colleges as a tool for identifying potentially unsuccessful students. Those identified students might then be aided by a counselor or other helping person to achieve academic success. Forewarned, they should be able to prepare themselves better to achieve academically. This aid
by a counselor to the individual student might be a stronger factor in his achieving academic success than all of the variables combined.
Further Research
There is a need for further research in the prediction of academic success. Specifically, the whole question of non-intellective variables and their contribution to the probability of academic success should be explored. Such variables as "local residence" while the student attends college, "marital status" at the beginning of college, and any changes that occur in this status, might be very important for educators to consider and need much more research. There are indications that these variables are more important in some colleges than others, but more research should be done.
A greater in-depth study of the individual colleges is needed. Such a study should look at additional variables such as composition of the population of the college with respect to social class background, socio-economic status, size of home town, part time employment of students, and others. This type of information is not available on student records but the review of related literature indicates that it might be valuable in the refinement of a predictive model for academic success.
- 56
Summary
Success predictions are but one of the many things to be considered during the process of exploration with a student while attempting to help him direct his academic life. The predictions should serve as tools in the exploration process.
As was suggested earlier by Justman (19), predictions should be tentative. Professional knowledge, judgment, and experience cannot be replaced by a predictive model for academic success, but can only be aided.
APPENDIXES
APPENDIX A
STUDENT REGULATIONS AT THE UNIVERSITY OF FLORIDA
Certain portions of the 1967 edition of The
University Record which pertain to probation, suspension,
and exclusion for academic reasons are presented.
PROBATION, SUSPENSION, AND EXCLUSION FOR ACADEMIC REASONS
The University of Florida is responsible for
providing the best possible education in an economical
and efficient manner. In order to discharge this
responsibility, the University must require reasonable academic progress from its students. Continuation of
students who have demonstrated a lack of the necessary ability, preparation, industry, or maturity to benefit
reasonably from a program of university study is
inconsistent with the University's responsibility as
a tax supported institution.
The University of Florida Senate has enacted
regulations covering probation, suspension, and exclusion. These regulations are directed toward enforcing the academic standards of the University. The academic standards of the University require both the maintenance
of grade point averages consistent with a reasonable chance of satisfactory completion of the University
programs and reasonable conformance to the catalog
description of the program of study in which the student
is engaged. Any college of the University may specify
additional academic standards and students are responsible
for observing the regulations pertaining to such
standards. (p. 127)
- 58 -
- 59
PROBATION BECAUSE OF UNSATISFACTORY
ACADEMIC ACHIEVEMENT
The purpose of academic probation is to recognize
formally the fact that a student is making unsatisfactory progress.
The conditions of academic probation are intended
to: (1) Relate to quality of achievement below standards required for ultimate graduation; (2) Recognize unsatisfactory work at an early date; (3) Be sufficiently significant to make clear to the student, his parents, and the administration, the shortcomings of the student's performance; (4) Provide occasion for counseling; (5) Give students whose ultimate success is doubtful further opportunity to demonstrate adequate performance.
AllZZ undergraduate students:
Any student who is eligible to return to the University after a suspension because of academic reasons (failure to receive passing grades in at least one-half of his work or having his load reduced to less than twelve hours), will be placed on scholarship probation for his next quarter.
In addition to University probation, a student may be placed on probation by the Cottllege in which he is registered if he does not maintain normal academic progress in the program of study in which he is engaged.
Upper division students:
An upper division student not on probation who fails to maintain a "C" (2.0) average for all work attempted in any quarter will be placed on scholarship probation for his next quarter. (p. 128)
CONTINUATION OF PROBATION
Upper division students:
An upper division student on scholarship probation who withdraws prior to the final date published in the
- 60
University Catalog will be continuted on probation if his average is above "C" (2.0) for all work attempted while registered in his present upper division college.
REMOVAL OF PROBATION
Upper division students:
Scholarship probation.will be removed for any full-time upper division student who earns a grade point average of "C" (2.0) or higher during the quarter when he is on scholarship probation at the University of Florida.
Removal of college probation:
A student will be removed from college probation when it is deemed by his college that he is making normal academic progress in the program of study in which he is engaged. (p. 128)
SUSPENSION
The purpose of suspension from the University for
academic reasons is to remove from the University those students who would not ultimately meet requirements for graduation if they continued at their current level of progress.
The conditions of academic suspensions are intended to: (1) Select students whose performance indicates that they will not fulfill the requirements for graduation; (2) Encourage students to leave the University as soon as a high probability of failure is evident.
AllZZ undergraduate students:
All undergraduate students (i.e.,students classified other than 7) who do not receive passing grades (A,B,C,D) in at least one-half of the hours attempted in any quarter shall be suspended immediately from the University. If such a student was not on probation for academic reasons and had not been previously suspended for
- 61
academic reasons, the suspension shall be for one quarter. If the student was on probation for academic reasons or had been previously suspended for academic reasons, he shall be excluded from the University without the opportunity to re-enroll. However, failure in only one course carrying five quarter credits or less shall not cause the student to be suspended under this provision.
All undergraduate students who are dropped from a course because of excessive absences or unsatisfactory work and as a result of such action are left with an academic load of less than twelve credits, shall be suspended immediately. If such a student was on probation or had been previously suspended for academic reasons, he shall be excluded from the University without the opportunity to re-enroll. If the student was not on probation or had not been previously suspended for academic reasons, the suspension shall be for one full quarter.
Any student who receives a second EW (dropped for non-attendance or unsatisfactory work) in military science courses will be suspended from the University for one full quarter.
Upper division students:
An upper division student who is on scholarship
probation will be ineligible for further registration at the University unless he maintains an average of "C" (2.0) in all work attempted that quarter or has an average of "C" (2.0) in all work attempted while registered in his present upper division college. (p. 129)
EXCLUSION
Exclusion from an undergraduate program of study:
A student may be excluded from a program of study by the College responsible for the program if he fails or refuses to maintain normal academic progress. Such exclusion does not prohibit the student from enrolling in other programs or colleges if he meets the requirements. (p. 130)
- 62
The student regulations were changed in the years
between 1967 and 1970. For that reason certain portions
of the 1970 edition of The University Record which pertain
to probation, suspension, and exclusion for academic
reasons are included.
PROBATION BECAUSE OF UNSATISFACTORY
ACADEMIC ACHIEVEMENT
The purpose of academic probation is to recognize
formally the fact that a student is making unsatisfactory progress.
The conditions of academic probation are intended
to: CI) Relate to quality of achievement below standards required for ultimate graduation; (2) Recognize
unsatisfactory work at an early date; (3) Be sufficiently significant to make clear to the student, his parents, and the administration, the shortcomings of the student's performance; (4) Provide occasion for counseling; (5) Give students whose ultimate success
is doubtful further opportunity to demonstrate adequate
performance.
AllZ undergracrduate students:
A student with less than a 2.0 grade point average
in his respective division (lower, upper) shall be
placed on scholarship warning if he has a grade point
deficit of nine or less.
A student with less than a 2.0 grade point average
in his respective division shall be placed on scholarship probation if he has a grade point deficit of ten
or more but less than twenty.
Any student who is eligible to return to the
University after a suspension because of academic
reasons will be placed on scholarship probation for
his next quarter.
In addition to University probation, a student may
be placed on probation by the College in which he is
- 63
registered if he does not maintain normal academic progress in the program of study in which he is engaged. (p. 161)
CONTINUATION OF PROBATION
AllZZ undergraduate students:
A student's scholarship warning shall be continued as long as he has a grade deficit of one but not more than nine. A student's scholarship probation shall be continued as long as he has a grade point deficit of ten but not more than nineteen. If his grade point deficit places him in another probation category, he shall be subject to the provisions of that category.
REMOVAL OF PROBATION
AllZZ undergraduate students:
Scholarship probation and/or scholarship warning will be removed when a student's grade point deficit in his respective division has been reduced to zero.
Removal of college probation:
A student will be removed from college probation when it is deemed by his college that he is making satisfactory academic progress in the program of study in which he is engaged.
SUSPENSION
The purpose of suspension from the University for
academic reasons is to remove from the University those students who would not ultimately meet requirements for graduation if they continued at their current level of progress.
The conditions of academic suspensions are intended to: (I) Select students whose performance indicates that they will not fulfill the requirements for graduation; (2) Encourage students to leave the University as soon as a high probability of failure is evident.
- 64
AllZ undergraduate students:
A student with a grade point deficit of twenty or more in his respective division (lower, upper) shall be suspended from the University for one quarter.
A student re-enrolling after a one quarter suspension will be on final scholarship probation, and if his grade point deficit is twenty or more at the end of the quarter he re-enrolls, he will be suspended without the possibility of re-registering except by committee action. (p. 162)
EXCLUSION
AllZZ undergraduate students:
A student may be excluded from a program of study by the College responsible for the program if he fails or refuses to maintain normal academic progress. Such exclusion does not prohibit the student from enrolling in other programs or colleges if he meets the requirements. (p. 163)
CLASSIFICATION OF STUDENTS-FLORIDA OR NON-FLORIDA
For the purpose of assessing fees, applicants shall be classified as Florida or non-Florida students. A Florida student is a person who shall be a citizen of the United States or a resident alien and who shall have resided and had his habitation, domicile, home and permanent abode in the State of Florida for at least twelve (12) months immediately preceding his current registration. In applying this regulation, "applicant" shall mean a student applying for admission to the institution if he is married or 21 years of age, or, if he is a minor, it shall mean parents, parent, or legal guardian of his or her person.
In all applications for admission by students as citizens of the State, the applicant, if married or 21 years of age, or, if a minor, his parents or legal guardian shall make and file with such application a
- 65
written statement under oath that such applicant is a bona fide citizen and resident of the State aid entitled as such to admission upon the terms and conditions prescribed for citizens and residents of the State.
In the determining of a Florida resident for
purposes of assessing fees, the burden of proof is on the applicant. Under the law an applicant can change his place of residence from another state to the State of Florida only by actually and physically coming into the State and establishing his residence with the intention of permanently residing within the State. The domicile or legal residence of the wife is that of the husband, and the legal residence of a minor is that of the parents, parent, or legal guardian of his person.
A non-Florida student may apply in writing for
reclassification prior to any subsequent registration under the provisions set forth below. To qualify for reclassification as a Florida student, a person (or if a minor, his parents) shall have resided in Florida for twelve (12) months, shall have filed a declaration of intent to become a resident of the State, and shall be registered to vote in the State. An alien shall have resided in Florida for twelve (12) months and must present U. S. Immigration and Naturalization certification that he is a resident alien. If the application is supported by evidence satisfactory to the University that the student then qualifies as a Florida student, his classification will be changed for future registrations. (p. 139)
APPENDIX B
COMPUTER CARD DATA CODING
Col. Item
02-10 Social Security Number 11-12 College Membership 01 if AS, 02 if BA, 03 if ED,
04 if JM
13 Academic Success 1 if graduated, blank if not 14-16 Lower Division G.P.A. (Freshman and Sophomore
years)
17-19 12th Grade Test Score or equivalent 20-22 Age in Months from birth to entry into Junior
year. Possible Junior starting times are 9/67,
1/68, 4/68, 6/68
23 Sex 1 if male, 2 if female 24 Native or Transfer 1 if native, 2 if transfer
(Native is less than 11 transfer hours)
25 Resident 4 if Florida, 1 if non-Florida,
3 if Alien
26 Residence 1 if Campus, 2 if Frat. or Sor.,
3 if Campus Married, 4 or 5 if Off Campus 27 Race 1 if White, 2 if Negro 28-30 Number of times on probation 31-33 Number of times student was part-time since becoming
a Junior
34-35 Number of term first part-time occurred (Fall 67=05
36-41 SCAT Scores
- 66 -
- 67
42-47 SAT Scores
48-51 Matriculation Information (Term by month and year,
status)
52-53 Institution Code (From which matriculated) 54 Marital Status as of first Junior status 1 if
single, 2 if married, 3 if divorced, 4 if widowed 55 Second marital status if different from above 56-57 Number of term in which first change in marital
status occurred
58-60 Transfer Hours
61-62 Terms enrolled while in Junior or Senior status 63-64 Term student entered his Junior year of study
(Fall 67=05, Winter 68=06. .) 65-80 Name of Student
BIBLIOGRAPHY
1. Anderson, Harry E., "Regression, Discriminant Analysis,
and a Standard Notation for Basic Statistics," in Handbook of Multivariate Experimental Psychology.
Edited by Raymond B. Cattell, Chicago: Rand McNally
and Company, 1966.
2. Atwell, Charles Alan, "Institutional and Community
Characteristics Related to the Effectiveness of
Transfer Programs in Florida Public Junior Colleges,"
Doctoral Dissertation, University of Florida, 1968.
3. Bashaw, Wilbur Louis, "A Central Prediction System for
Predicting the Success of Junior College Transfers in Florida Universities," Doctoral Dissertation, Florida
State University, 1963 (Dissertation Abstracts, Vol.
25, p. 282).
4. Bauer, Roger, William A. Mehrens and John F. Vinsonhaler,
"Predicting Performance in a Computer Programming
Course," Educational and Psychological Measurement,
1968, Vol. 28, No. 4, pp. 1159-1164.
5. Bayes, Andrew Hartin, "An Application of Hotelling's
Canonical Correlation to Academic Prediction,"
Doctoral Dissertation, University of Miami, 1968
(Dissertation Abstracts, Vol. 29, p. 2512).
6. Bottenberg, Robert A., and Joe H. Ward, Jr., Applied
Multiple Linear Regression. Technical I)ocumentary Report PRL-TDR-63-6. United States Department of Commerce. Washington: U. S. Government Printing
Office, 1963.
7. Byron, Anthony R., "Non-Intellective Variables Related
to Successful and Unsuccessful Students in a Junior
College," University of Missouri, 1968. ERIC
Microfiche ED 023 387.
8. Clarke, Johnnie R., and Rose Mary Ammons, "Identification
and Diagnosis of Disadvantaged Students," Junior College Journal, 1970, Vol. 40, No. 5, pp. 13-17.
9. Conklin, R. C., and D. G. Ogston, "Prediction of Academic
Success for Freshmen at the University of Calgary,"
- 68 -
- 69
Alberta Journal of Educational Research, 1968, Vol.
14, No. 3, pp. 185-192.
10. Cooley, William W. and Paul R. Lohnes, Multivariate Procedures for the Behavioral Sciences. New York:
John Wiley and Sons, Inc., 1962.
11. Cooper, Leland Ross, "The Relationship of Selected Factors to the Continuance of Junior College
Graduates at Senior Institutions," Doctoral Dissertation, University of Florida, 1964.
12. Dixon, W. T., editor, BMD: Biomedical Computer Programs, Berkeley and Los Angeles: University of California
Press, 1968.
13. Fishman, Joshua A., "Some Social-Psychological Theory for Selecting and Guiding College Students," The
American College. Edited by Nevitt Sandford, New
York: John Wiley and Sons, Inc., 1962, Part V, pp.
666-689.
14. Gadzella, Bernadette, and Grace Bentall, "Differences in High School Academic Achievements and Mental
Abilities of College Graduates and College Drop-Outs,"
College and University, 1967, Vol. 42, No. 3, pp.
351-356.
15. Hopper, Harold H., Predictors of College Success, 1968,
ERIC Microfiche ED 024 374.
16. Iffert, Robert E., Retention and Withdrawal of College Students. United States Office of Education Bulletin No. 1. Washington: U. S. Government Printing Office,
1968.
17. Irvine, Donald W., "Graduation and Withdrawal: An Eight-Year Follow-up," College and University, 1965,
Vol. 41, No. 1, pp. 32-40.
18. Ivanoff, John M., "The Use of Discriminant Analysis for Predicting Freshman Probationary Students at
One Midwestern University," Educational and Psychological Measurement, 1961, Vol. 21, No. 4, pp.
975-985.
19. Justman, Joseph, "The Counselor's Use of Measurement in Prediction," National Catholic Guidance Conference Journal, 1968, Vol. 12, No. 2, pp. 145-153.
- 70
20. Kerlinger, Fred N., Foundations of Behavioral Research.
New York: Holt, Rinehart and Winston, Inc., 1964.
21. Lavin, David E., The Prediction of Academic Performance.
New York: Russell Sage Foundation, 1965.
22. Magoon, Thomas M., and Martha J. Maxwell, "Demographic
Differences Between High and Low Achieving University Students," The Journal of College Student
Personnel, 1965, Vol. 6, pp. 367-373.
23. Maxwell, A. E., "Canonical Variate Analysis When the
Variables Are Dichotomous," Educational and Psychological Measurement, 1961, Vol. 21, No. 2,
pp. 259-271.
24. Ostle, Bernard, Statistics in Research. Ames: The
Iowa State University Press, 1963.
25. Renetzky, Alvin, editor, Yearbook of Higher Education1969. Los Angeles: Academic Media, Inc., 1969.
26. Richards, James M., Jr., John L. Holland, and Sandra
W. Lutz, "Prediction of Student Accomplishment in College," Journal of Educational Psychology, 1967,
Vol. 58, No. 6, pp. 343-355.
27. Rose, H. A., and C. F. Elton, "Another Look at the
College Drop-out," Journal of Counseling Psychology, 1966, Vol. 13, pp. 242-245.
28. Roudabush, Glenn E., "A Study in Prediction from
Biographical Information," Doctoral Dissertation,
University of Washington, 1963 (Dissertation
Abstracts, Vol. 25, p. 1324.
. 29. Schroeder, Wayne L., and George W. Sledge, "Factors
Related to Academic Success," The Journal of College
Student Personnel, 1966, Vol. 7, No. 2, pp. 97-104.
30. Sims, David M., "A Study of the Relationship of Selected
Institutional Characteristics of the Junior College
of Origin to the Academic Performance of Public
Junior College Transfer Students in the University System of Florida," Doctoral Dissertation, Florida
State University, 1966. ERIC Microfiche ED 026 241.
- 71
31. Summerskill, John, "Dropouts from College," The
American College, Edited by Nevitt Sandford, New
York: John Wiley and Sons, Inc., 1962, Part V,
pp. 627-657.
32. The University Record, Undergraduate Catalog Issue,
Gainesville: University of Florida, 1967.
33. The University Record, Undergraduate Catalog Issue,
Gainesville: University of Florida, 1970.
34. Thornton, James W., Jr., The Community Junior College,
2nd Edition. New York: John Wiley and Sons, Inc.,
1966.
35. Trent, James W., and Leland L. Medsker, Beyond High
School. San Francisco: Jossey-Bass, Inc., Publishers, 1968.
36. Walker, John E., Academic Performance of Native and
Transfer Students in the Upper Division of the University of Florida, 1966-1968. Gainesville:
Institute of Higher Education, 1969.
37. Woodward, Ivor Carey, "The Relative Efficiency of
Multiple Regression Analysis and Multiple Cutoff
Analysis in the Prediction of Academic Performance
in a Selected Medical School," Doctoral Dissertation, University of Southern California, 1968
(Dissertation Abstracts, Vol. 29, p. 1143).
38. Wyatt, Woodrow, and Charles M. Bridges, Jr., Statistics
for the Behavioral Sciences. Boston: D. C. Heath
and Company, 1966.
BIOGRAPHICAL SKETCH
Ronald Wheeler McFaddin was born August 31, 1934, at Cincinnati, Ohio. He attended elementary school in San Diego, California, and graduated from Manatee County High School in Bradenton, Florida, in 1952. He attended the University of Florida from 1952 to 1954 when he entered the Air Force and served as a pilot and administrative officer. He was released from military service in 1958 and took employment in sales and management. Returning to the University of Florida in 1965 he resumed studies and received the Bachelor of Science degree in Health Education in 1967. He enrolled in the Graduate School of the University of Florida where he studied counseling and guidance. He received the Master of Education degree in 1968 with a major in Personnel Services, and since that date has been studying toward the Doctor of Education degree with a major in Educational Administration with emphasis on higher education.
Ronald Wheeler McFaddin is married to the former Mary Katharine Moore and is the father of two children, David and Laurie. He is a member of Phi Delta Kappa and Kappa Delta Pi.
- 72 -
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Education.
J4 hes L. Wattenbarger, hairman Pfofessor of Educationd1
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Education.
Bert ISha
Professor Education
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Education.
Dyi n/ RoberyA
Associate Professor of Education
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the de Docto f Education.
FF 0 /ard
Assistant Proffessor of Education
I certify that I have read this study and that in
my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Education.
John James ,
Associate Professb of Management
This dissertation was submitted to the Dean of the College of Education and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Education.
March 1971
Dean, 2leggof Education
Dean, Graduate School
|
Full Text |
PAGE 1
A PREDICTIVE MODEL FOR ACADEMIC SUCCESS AT THE UNIVERSITY OF FLORIDA By Ronald Wheeler McFadclin A Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Education UNIVERSITY OF FLORIDA 1971
PAGE 2
Copyright by Ronald Wheeler Mcfaddin 1971
PAGE 3
ACKNOWLEDGMENTS The writer wishes to acknowledge the assistance of the many persons who took an active interest in the preparation of this study. The counsel and assistance of Dr. James L. Wattenbarger, Chairman of the writer's Advisory Committee, are deeply appreciated. Sincere thanks are also extended to the members of the committee, Dr. Bert L. Sharp, Dr. Dayton Y. Roberts, Dr. Robert T. Frossard, and Dr. John H. James. Appreciation is expressed to the many individuals of the Office-of Academic Affairs, especially Mr. Victor Yellen and Mr. Arthur Haller, the Office of Student Development, and the Board of University Examiners for their support and assistance in collecting data for this study. The writer is also indebted to Dr. Charles M. Bridges, Dr. Gerald R. Boardman, Mr. Gene A. Barlow, and Mr. William Graves for their advice and assistance during the study. To his wife, Katharine, his son,David, and his daughter,Laurie, the writer wishes to express his deepest gratitude for their love, devotion, patience, understanding, and assistance during the preparation of this study. iii
PAGE 4
TABLE OF CONTENTS ACKNOWLEDGMENTS ABSTRACT Pa ge iii vi Chapte r I. INTRODU C TION ............................... 1 The PP o b l em . . . . . . . 3 Assumption . . . . . . . 6 Definition of Te1 ms . . . . . 6 Review of Related Literature ..... ... ..... 8 Procedures . . . . . . . 16 Summary . . . . . . . . 18 II. IDENTIFYING THE DATA .. .. . . .. . 20 Introduction . . . . . . . 20 Collection of Data . . . . . 20 Computer Data Cards ...................... 23 Analysis of Data .. .. . . . .. 24 Summary . . . . . . . . 2 7 III. DEVELOPMENT OF THE MODEL .................... 28 Introduction . . . . . . . 28 Developing the Model . . . . . 2 8 Arts and Sciences . . . . . . 32 Business Administration .................. 36 Educa:ti e n . . . . . . . . 4 0 Journalism . . . . . . . 44 Demonstration of the Predictive Model . . . . . . . . 48 Summa r y . . . . . . . . SO IV. IMPLICATIONS AND RECOMMENDATIONS . . 52 In trodu.e tio n . . . . . . . S 2 Implicat,Z:o n s .'............................ 52 Weakness e s of the M odel .................. 53 JV
PAGE 5
Chapter IV. Appendix A. B. TABLE OF CONTENTS~Continued Use of the Predictive Model ............ Further Research ....................... Summary ................................ STUDENT REGULATIONS AT THE UNIVERSITY OF FLORIDA ............................. COMPUTER CARD DATA CODING ................. BIBLIOGRAPHY BIOGRAPHICAL SKETCH ... ........................... V Page 54 55 56 58 66 68 72
PAGE 6
Abstract o f Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Education A PREDICTIVE MODEL FOR ACADEMIC SUCCESS AT THE UNIVERSITY OF FLORIDA By Ronald Wheeler Mcfaddin March, 1971 Chairman: Dr. James L. Wattenbarger Major Department: Educational Administration The high attrition rate of students who enroll in institutions of higher education has raised serious questions concerning the responsibility of higher educational institutions to increase the probability of academic success for these students. This study attempts to dea l with one of these responsibilities, that of guidance, throug h the d evelopment of a predictive mode l for academic success. Students enrolled in four colleges of the University of Florida as first-time juniors duri ng the period Fall 1967 through Summer 1968 and who met the criterion of academic success o r its antithesis were the s ample for this study. The sample include d 2,259 students. The term academic success was d efine d a s the attainment of the bach elor's VJ.
PAGE 7
degree from the college entered at the beginning of the junior year, and its antithesis was the withdrawal of the student from the college while under academic probation or suspension. Academic records of the sample students were followed from their term of entry through Summer 1970. The data collected were taken from existing student records maintained by the University of Florida, and no new data were created for this study. The data were subjected to stepwise discriminant anaiysis by a computer technique of the BMD Biomedical Computer Programs package. Four sets of models were produced (one set for each college) which may be used for the prediction of academic success or its antithesis. The program, on the basis of its developed model, predicted success or withdrawal for each of the students in the sample. Based on the actual performance of these students these predictions misclassified a number of cases. The percentages of misclassification ranged from 20 to 28 per cent for the four colleges. The findings indicate that the probability of a student achieving academic success at the University of Florida might be directly related to his choice of college enrollment within the university. People who counsel students in their academic lives need to be aware of such possible different outcomes between the colleges within the institution. vii
PAGE 8
Those who counsel students should be cognizant that a predictive model is a tool. Professional knowledge, judgment, and experience cannot be replaced by a predictive model for academic success, but can only be aided. ) viii
PAGE 9
CHAPTER I INTRODUCTION Of the students who entered higher educational institutions over the past several decades, approximately 60 per cent eventually were graduated from some institution of higher education (17). Of the 40 per cent who did not graduate, how many could have been saved had administrators, counselors, or teachers known, through the use of some academic success predictor model, the probability of academic success for these students? Had these students been forewarned with individualized guidance, and thereby better able to prepare themselves, would the percentage of graduates have been much higher? It is the responsibility of higher educational institutions to become aware, on individual bases, of high risk students so that individual, special efforts can be made to reduce the probability of failure for these students to succeed academically. Schroeder and Sledge studied factors related to collegiate academic success. In their review of recent studies, they found those studies in the minority which used specific collegiate averages and/or averages acquired beyond the first year of college. They also found that methodology was restricte d primarily to simple and multiple -1 -
PAGE 10
-2 -correlations supplemented, in some cases, by tests of significance. They reported that intellective factors were found to be more predictive of collegiate achievement than non-intellective factors although the importance of the latter was not disputed (2~). Schroeder and Sledge reported that multiple correlation coefficients, computed in an effort to determine the best possible predictive combinations, did not exceed .71. Some studies, however, by utilizing freshman grades as one of the predictors of achievement beyond the freshman year, were able to secure a multiple correlation beyond .80. (29, p. 98) Such high correl~tions would be very valuable to a counselor as predictors. The present study used both intellective and non-intellective factors as predictors of collegiate achievement. Freshman and sophomore grades were considered among the intellective factors in this study. There is a need for a model at the University of Florida to predict the probability of academic success on the basis of common data available on student records. A counselor or other helping person would then be able to work individually with each potentially unsuccessful student to increase his probability of academic success. This study used a multivariate analysis computer program technique in the attempt to develop such a model to predict academic success at the University of Florida.
PAGE 11
-3 -The Problem Statement of the Problem The purp6se of this study was to develop a set of models for predicting academic success for students in selected colleges at the University of Florida. Justification for the Study Although attempts have been made over the years to provide models for the prediction of academic success, only the recent advent of high-speed computers and the concomitant development of sophisticated techniques of data treatment have made this topic practicably researchable (6). The problem for this study was a practical problem, as many administrators and counselors have been, and will be, faced by students with the question: What chance do I have? The development of a predictor model would provide a valuable tool for people who help and guide the academic lives of students. If facts, predicated by such a model, are presented to the student, it could be made quite clear to him just how much his own desire to succeed would really count. Often it is this desire, this motivation, on the part of the student which causes the difference between academic success and academic failure. The Unite d States needs as many of its students to succeed academically as arc able. T}1e normal 40 per cent who
PAGE 12
-4 -are lost represents a tragic waste which this nation cannot afford. If the 1967 Fall enrollment of 1,652,317 first-time freshmen in all public and private institutions in the United States were considered, the latest year reported by the OEO as reported in the Yearbook of Higher Education~1969, this would amount to nearly 661,000 youth lost (25, p. 382). In research related to the junior college, it was suggested by Thornton (34) and by Trent and Medsker (35) that for 35 per cent to 70 per cent of the students who are not academically successful,. their failure was to some extent due to a failure of the college. One of the major reasons given for failures was the lack of adequate guidance for students, and that generalization might be found to be true in senior institutions too. A predictor model for academic success could be a valuable tool in providing more adequate guidance. At the present time, there is no tool being used at the University of Florida which considers the many variables included in this study. Delimitations and Limitations Delimitations: 1. This investigation was confined to students in the junior and senior years in the colleges of Arts and Sciences,
PAGE 13
-5 -Business Administration, Education, and Journalism at the University of Florida. 2. Only data made available to the researcher by the Office of Academic Affairs of the University of Florida were utilized in this study. The data available were limited to the period Fall 1967 through Summer 1970. 3. This investigation was limited to those variables indicated in the research design, namely: academic grade point average for freshman and sophomore years, Florida Twelfth Grade Test score or its equivalent, age of student in months, from birth to entry into his junior year, sex of student, native or transfer student, Florida or non-Florida resident, local residence, race of student, and marital status of student at the beginning of his junior year of study. 4. The analysis of data was confined to the stepwise discriminant analyses of the investigated colleges by a computer technique of the BMD Biomedical Computer Programs package which produced a set of equations from which a set of predictor models of academic success could be evolved (12). Limitations: This study was an ex post facto research study and subject to the limitations of this method (20). It was also subject to the degree of the reliability of the records use d to collect the d ata. Certain information which may be
PAGE 14
-6 -considered as predictors, e.g., a measurement of motivation, wa~ not available to the researcher for this study. Assumption It was assumed for the purpose of this study that like grades are equal, e.g., the grades of "A" are equal regardless of institutional, college, or professorial source. Definition of Terms The basic documents for the definition of terms were The University Record, of the University of Florida, Under. graduate Catalog Issues for the years 1967 and 1970 (32, 33). Academic probation~That condition of formal recognition of the fact that a student is making unsatisfactory progress. For the purpose of this study, this term is inclusive of all degrees of academic probation as defined in the Undergraduate Catalogs (32, 33). For a complete explanation of this term see Appendix A. Academic success~For the purpose of this study, this term will mean the attainment of the bachelor's degree from the college entered at the beginning of the junior year of study. This is a narrow definition of academic success, but the antithesis will also be narrow in that this group will include only those students who withdrew from their college while under academic probation or were suspended. The large group of students who might have withdrawn from the institution to attend another, or transferred to another college, or continued in attendance beyond the period of this study, or withdrew for any of a number of other reasons, were considered to be neither success nor failure; they were excluded from this study. Academic withdrawaZ~For the purpose of this study, this term will be the antithesis of the term Academic Success. Entry age~The age of the student in months, calculated from date of birth to date of entry in his junior year of study.
PAGE 15
-7 Florida r esident~A person who is a citizen of the United States or a resident alien and who shall have resided and had his habitation, domicile, home and permanent abode in the State of Florida for at least twelve months immediately preceding his current registration (32, 33). Non-Florida resident includes all those students who do not meet the resident requirements as listed above. For a complete explanation of this term, see Appendix A. Florida Twelfth Grade Test score~That score or its equivalent as converted from an acceptable general ability test, e.g., SAT or SCAT, which ranks the high school seniors of the State of Florida. The official name of this program is the Florida State-Wide Twelfth Grade Testing Program. Grade point average~The average as determined by computing the ratio of grade points to credit hours attempted. Grade points are established by equating each credit hour as follows: A with 4, B with 3, C with 2, D with 1, and E, EW, I, WF, and X with 0. Intellective factors~Those variables which are measured by ability, aptitude, or achievement tests, or some measure of prior achievement such as grade point average. Junior year of study~For the purpose of this study, this term will include only those students who enter one of the included colleges for the first time and, by the University of Florida, are classified as 3AS, 3BA, 3ED, or 3JM. Model~The models developed in this dissertation are represented by regression equations in which the weighted inde pendent variables selected from the list above will predict the probability of the dependent variable, academic success. These equations were developed by using a BMD Program computer technique. Native student~The student admitted to the Upper Division from the University College of the University of Florida who has taken not more than 10 credit hours from another institution. The limitation of 10 credit hours was also used by Walker in his study (36). This allows for the student who takes a course or two during the summer at another institution closer to home still to b e considered a native student.
PAGE 16
-8 Non-intellective factors~Those variables not included in the term intellective factors, e.g., entry age of student for junior year, sex of student, native or transfer student, marital status of student at beginning of junior year, Florida or non-Florida resident, on or off campus residence, and race of student. Suspension and exclusion--For the purpose of this study, these terms may be used interchangeably. As stated in The University Record, of the University of Florida, the purpose of suspension from the University for academic reasons is to remove from the University those students who would not ultimately meet requirements for graduation if they continue at their current level of progress (33). For a complete explanation of these terms see Appendix A. Transfer student~That student admitted to the Upper Division from any source other than the University College of the University of Florida, and those students admitted from the University College who transfer more than 10 credit hours from other institutions. This group includes students from public and private junior colleges, students from public and private senior institutions of Florida, and students from out-of-state senior institutions. Upper division~The junior and senior years of the colleges included in this study. Review of Related Literature The development of a predictor model for academic success was drawn from a review of such questions as: the college attrition problem, methods of studying the attrition problem, and types of analysis of those methods. A report of the review of these questions is presented in this section. The Colle g e Attrition Pr oblem An investigation by Cooper cited the results of a governmen t a l study conduc t e d in 1938 by McNeely which dealt
PAGE 17
-9 -with data collected from over fifteen thousand students in 25 universities. The attrition rate at the end of four years was found to be 62.1 per cent. He reported a number of students later enrolled in the same or another school and eventually received a degree, reducing the net attrition rate to 45.2 per cent (11). A study reported by Iffert in 1958, with a sample of institutions, found that slightly fewer than 40 per cent of students entering the higher educational institutions in the study graduated from the institution of original registration in normal progression~that is, in a 4-year period~ and perhaps 51 per ~ent graduated from some institution in this period. The conclusion that nearly 60 per cent of the students in the study eventually graduated from some institution of higher education seems to be justified by the data. (16, p. 20) Summerskill reported in 1962 that he reviewed 35 different studies that cited attrition rates for classes entering hundreds of varied colleges and universities from 1913 through 1962. He concluded that apparently the attrition rate had not changed appreciably during that 40-year period. While the rates varied from 12 per cent to 82 per cent in the 35 studies, about 40 per cent of college students, over the past several decades, have graduated on schedule and perhaps another 20 per cent graduated from some college at a later time (31). Irvine reported tha t 49.S per cent of entering freshmen at the University of Georgia in 1955 had graduated by
PAGE 18
-10 -1963 fiom some institution. This was a period of eight years. It was found that women more frequently than men graduated "on time," but that in the long run, slightly more men obtained degrees from so~e college (18). Walker studied transfer and native students in the upper division of the University of Florida. He reported thit 62.S per cent of native students and 32.2 per cent of transfer students who had entered the upper division in the fall term of 1966 had graduated by the end of the summer term of 1968 (36). The implication is that there may be sharp differenc~s in aitrition rates among different groups of a student body. We may conjecture that the more restrictive admission policies since the mid-sixties have changed the attrition trend of the past several decades. However, there is no evidence in the literature to support such an idea. This nation cannot afford this continued waste of human talent, and it is fitting that efforts be made in the decade of the seventies which will improve the record by providing tools for predicting academic success and helping students make decisions appropriate to their needs. The high-speed computer will provide the ability for this undertaking to allow counselors and administrators a better understanding of individual students.
PAGE 19
-11 -Methods of Studying the Attrition ProbZem Atwell reported that early investigators of prediction usually examined the relationship of some intellective variable, e.g., an aptitude or achievement test score or some measure of prior achievement such as grade point average or rank-in-class, to a criteria~ or dependent variable which was nearly always some type of grade average. Later studies have included non-intellective factors in the research design and have often used a multi-statistical technique to measure the combined predictive value of two or more variables (2). In addition to this type of study, other investigators have concentrated on non-intellective factors. Roudabush concluded that biographical information can contribute substantially to the prediction of academic criteria (28). The present investigation used non~intellective, biographical information as some of the variables studied. Richards ~tat. used non-academic accomplishment such as participation in extra-curricular activities as predictors. These were found, however, to be largely independent of academic potential and achievement (26). Research using nonintellective variables also was stressed by Lavin in his book on predicting academic performance (21). One of the more concise discussions of this subject was by Fishman in which he criticized research people for
PAGE 20
\ 12 simply adding non-intellective measures to the already existing test batteries. These additions were usually a personality test type instrument. Fishman argued that these instruments seemed to be ''measuring something insufficiently dissimilar from whatever it is that our usual .predictors are measuring'' (13, p. 670). He explained his reasoning for taking this stand by a discussion of the usual predictive research method where the most usual predictors are high school grades and scores on a standardized measure of scholastic aptitude. The usual criterion is the freshman average. The average multiple correlation is approximately .55. The gain in the multiple correlation upon adding a personality test score to one or both of the usual predictors, holding the criterion constant, is usually less than .OS. (13, p. 699) An example of a study which used non-intellective variables, but of a discrete ~ature, ratber than continuous, is that by Byron. He attempted to distinguish between successful and unsuccessful junior college students on the_ basis of such non-intellective variables as: residency (whether or not the student lived in the college district), age, sex, marital status, on or off campus housing occupancy, part time work, and others. His study found only sex and part time work to be significantly different for the two groups. He found that if the student were female and worked part time, the chances of being successful were greater. He recommended further investigation of non-intellective variables to increase the knowledge of student
PAGE 21
13 behavior (7). The present study has included most of these same non-intellective variables. Types of Analysis Hopper used multiple regression in his study and the results generally supported the efficacy of considering grades when one wishes to predict grade success at the next level. But, more importantly, this study resulted in some rather substantial multiple R's, and introduced some non-intellective variables that had not heretofore been studied in any frequency (15). Those variables included study habits and attitude tests. His large multiple R's were done with Florida Twelfth Grade Test Scores. A study by Rose and Elton concerned an investigation of personality differences between groups of students who withdrew from colleges at different times. They used multiple discriminant analysis which revealed significant differences among these groups of students. This was with respect to certain personality traits (27). A study of demographic differences by Magoon and Maxwell, which made use of Chi-square test techniques, found significant differences within the colleges studied, but not between the coileges studied on the variables tested (22). Bayes' review of the literature showed that the best prediction of subsequent academic performance is r
PAGE 22
14 obtained from multiple correlation 1n which a battery of intellective variables are used to predict the overall grade point average. He found that canonical correlation does not provide more adequate prediction of academic achievement, nor is it more logically defensible than multiple regression in the prediction of the academic achievement of college freshmen. His study also showed that personality variables do not significantly increase the percentage of predicted variance in academic prediction cs) It was shown by Maxwell that discriminant function and canonical variate analysis can be applied to a problem involving dichotomous variables (23). Woodward used both continuous variables and dichotomous data in his study for prediction of academic performance (3 7). The pre s ent s tudy also use d both continuous and dichotomous d a t a i n the searc h for a predictor mod el. An instrument was produced by Ivanoff using discriminant analysis and continuous data for predicting freshman probationary students. Prediction tables were constructed for male and female students to serve as tools for personne l workers while helping students (18). In a study by Conklin and Ogston, a selection of achievement, intellective and persona l ity tests w e r e administe r e d to 639 u niversity freshme n in an atte mpt to
PAGE 23
-15 -identify variables related to first year success. Correlation and regression analyses were conducted on the data. The results showed high school averages to be the best predictor while the other variables were shown to possess little predictive utility (9). The study by Gadzella and Bentall produced much the same results though with different groups (14). Sims made his study in Florida and was able to conclude that lower division grade point averages and scores from the Florida Twelfth Grade Testing Program were significant predictors of upper division grades (30). The study by Bauer et aZ. reached a similar conclusion in that they found the best single predictor of academic success to be past academic success (4). The present study included these same variables in the production of the predictor model. A recent study by Clarke and Ammons considered disadvantaged students. They found that selected intellective instruments are predictors of academic success for white males and females, but none of their cognitive measures proved to be significant for Negro males (8). The variable of race was considered jn the present study; Bashaw developed a "central prediction system" to forecast the success of junior college transfers in Florida universities. He considered intellective variables,
PAGE 24
16 -and no nonintellective variables in his study (3). The present study did consider intellective and nonintellective variables as well as transfer and native students as variables in developing a model for predicting academic success. Justman reported that both precise and imprecise measurements are used by counselor personnel. He cautioned that, because of the many controllable factors involved, predictions should be tentativ~ Predictions should be based on recent evidence which is relevant to the area of prediction, and should grow out of conside~ation of as many independent bases ~s are available. He added that, in the end, subjective evaluation will of course play a prominent part (19). However, it would be well to have a predictor tool as part of the evidence available. The development of such a predictor tool was the purpose of this present study. Procedures Study D esign This was an ex post facto research study in which several continuous and discrete independent variables obtained from records of students were observed. From these, the dependent variable of academic success was predicted.
PAGE 25
-17 -The purpose of this design was to lead to the development of sets of mathematical models built upon these independent variables. Such models could be useful as a tool in predicting academic success and guiding students' academic lives. Sample The sample for this study was all students who entered the colleges of Arts and Sciences, Business Administration, Education, and Journalism at the University of Florida in their junior year of study during the first academic year of the period of this study, and who met the criterion of Academic Success or its antithesis as defined above. Data Collection Data on each student included in this study were supplied by the Office of Academic Affairs of the University of Florida. The data included most of the variables considered by previous researchers to be the "best predictors" of academic success for college students. The data included in this study were as follows: 1. Lower division grade point average (freshman and sophomore years) 2. Florida Twelfth Grade Test score or its equivalent 3. Age of student at beginning of junior year
PAGE 26
18 4. Sex of student S. Native or transfer student 6. Florida or non-Florida resident 7. Local residence 8. Race of student 9. Marital status of student at beiinning of the junior year Data Treatment The data collected were subjected to the stepwise discriminant analyses by a computer technique of the BMD Biomedical Computer Programs package. The College of Education Research Service Center has adapted these procidures to the University's IBM 360 computer for application in educational research. The data for each college were given separate treatment. The result of the use of the treatment technique was the development of a prediction model for academic success for each college included in this study. The models show a weight factor for each of the variables included, ranking these from most significant to least significant. Summary From the review of related literature it appears that almost half of the students who enter institutions
PAGE 27
19 of higher education fail to complete their studies and receive a baccalaureate degree. Additional information, such as a predictor model for academic success, may help more students to succeed. Previous research emphasis has b een based on the assumption that academic success predicts academic success. While this study considers this point, it goes beyond and attempts to consider other variables that may be indicative. Recent concern for disadvantaged students and others who have not achieved academic success has raised the question of relying solely on this assumption and pointed out that we require better answers than have been previously found if we are to properly do our jobs as educators. Chapter I is an introduction to this study which outlines the problem and includes a review of the related literature. Chaper II describes the procedure used to develop the model for the prediction. Chapter III is an analysis of the data resulting from the study and an outline of the development of the model. Chapter IV introduces some theoretical considerations and implications and suggests use of the predictor model for academic success.
PAGE 28
CHAPTER II IDENTIFYING THE DATA Introduction This was an ex post facto research study in which several continuous and discrete independent variables obtained from records of students were observed. From these independent variables, the dependent variable of academic success was predicted. The purpose of this study was to develop a set.of mathematical models (the predictive model) for predicting academic success for students in selected colleges at the University of Florida. coiiection of Data Data collection for this study was made possible by the Office of Academic Affairs of the University of Florida. The design of the study required examination of complete academic records of all the students who entered each of the four colleges included in the study as juniors during the period of Fall 1967 through Summer 1968. These colleges were Arts and Sciences, Business Administration, Education, and Journalism. The data utilized in this study were take11 from existing, individual student records as maintaine d by the 20 -
PAGE 29
-21 -Registrar and Admissions Office of the University of Florida. All of the information headings maintained on these records were studied by the researcher and the selection of those included in the study was based on data used by previous researchers as cited in the review of the literature, and the commonality of availability. It is important to note that no new data were created for this study. All data used were already available on student records maintained by the University of Florida. Often many of the records of transfer students do not include as much information as the records of native students. It is important that any data used in the development of a predictor model be commonly available on student records. The inclusion of variables which are commonly found to be missing would limit the usefulness of any success predictor model. In addition, certain kinds of information that might be useful in a predictor model, such as the record of part time employment by a student, simply are not maintained on these records. The population for this study was compiled from the cumulative tapes of student records of the Registrar and Admissions Office. The computer print-out of this compiled population included the academic record of each individual student fro~ his time of entry at the beginning of this study through the term of Summe r 1970. It was
PAGE 30
-22 -possible to observe individual progress by academic term, including end of term actions taken regarding each student, for a maximum of twelve academic terms. As these academic terms were academic quarters, normal progress would have been a total of six academic terms to graduation, that is, the completion of the junior and senior years of college. This population included 3,156 students. Inspection of this population led to the selection of the study sample. Inspection was made on the basis of the criterion Academic Success or its antithesis, Academic Withdrawal, as defined above. This produced a sample of 2,344 students who had met these qualifications. A print-out of the complete record of each of these students was extracted by computer from the various cumulative tapes maintained by the Office of Academic Affairs and the Registrar and Admissions Office. These data showed specific information to be missing which were necessary to this study. The largest amount of missing data was in the Florida Twelfth Grade Test score section. Also high in the miss-ing area was the section for grade point average for freshman and sophomore years. It should be noted that these data were only missing from the cumulative tapes and not the individual student records. Because of the missing data, an individual records search was necessary for approximately 1,000 students.
PAGE 31
23 This examination of individual student records was made possible by the Office of Student Development which allowed the researcher access to their files. Only eighty-five students who were included in the original sample could not ultimately be included in the study because of the lack of information on their individual records. This left a final sample of 2,259 students for the study. Over 200 Florida Twelfth Grade Test score equivalents were provided by the Board of University Examiners for ACT, SAT, and SCAT scores for individual students. The Florida Twelfth Grade Test was created by the Board of University Examiners and comparative studies by them over the years have enabled them to make up equivalent scores for these other examinations which are statistically reliable. This conversion wa~ necessary for standardization so that these individuals could be included in the study a5. defined above. Computer Data Cards Data cards were keypunched for each student and included the following information for this study: 1. Lower division grade point average (freshman and sophomore years) 2. Florida Twelfth Grade Test score or equivalent 3. Age of student in months, from birth to entry into junior year (possible starting dates for the junior year were: 9/67, 1/68, 4/68, or 6/68)
PAGE 32
-24 4. Sex of student 5. Native or transfer student (native students have less than 11 transfer hours of credit) 6. Resident (Florida, non-Florida, or alien) 7. Residence (campus, fraternity or sorority, campusmarried housing, or off campus) 8. Race of student (white or Negro) 9. Marital status of student at beginning of the junior year of study (single, married, divorced, or widowed) Additional information was keypunched, some of which was needed for identification purposes during the analysis of the data, such as: 1. College membership (Arts and Sciences, Business Administration, Education, or Journalism) 2. Academic success or its antithesis, academic withdrawal 3. Social security number See Appendix B for the coding of the data cards. Analysis of Data The method of analysis of these data was a multivariate analysis of the BMD Biomedical Computer Programs package developed at the University of California (12). The particular program used was identified as BMD07M Stepwise Discriminant Analysis. This program was designed for use where data are categorized by case and the cases are designated as belonging to one of two or more groups. Linear sums of the variables are determined
PAGE 33
-25 -which classify the cases into groups. This program proceeds in a stepwise manner by forming linear sums of first one, then two, three, etc., variables. At each step the variable added is the one which gives the greatest improvement in classification. The variable entered is selected by the first of the following equivalent criteria: 1. The variable with the largest F value. (This statistic is the Fisher F. The use of the F value in this procedure gives the largest likelihood of difference with all the variables in the equation in order to maximize the difference in the groups.) 2. The variable which when partialed on the previously entered variables has the highest multiple correlation with the groups. 3. The variable which gives the greatest decrease in the ratio of within to the total generalized variances. For a more complete description of this program see Dixon (12, p. 214a -214b). The problem in discriminant analysis is getting a linear combination of variables which distinguish, better than any other linear combination, between-the two or more chosen groups. Descriptions of discriminant function analysis can be found in Anderson (1), and Cooley and Lohnes (10). Discriminant analysis is very similar in procedure to multiple regression analysis. The principal difference between the two techniques lies in the nature of the d ependent or criterion variable In regression analysis, the dependent variable is a continuous variable taking
PAGE 34
-26 -infinitely many values; e.g., grade point average. In discriminant analysis, the dependent or criterion variable is group membership, which may or may not be on a continuum (it may be discrete). Thus discriminant analysis begins with defined groups and seeks to weight scores in order to achieve a maximal difference among group means; e.g., with the previously defined groups Academic Success and Academic Withdrawal. The data were separated by college so that four sets of models might be produced, one for each college in the study. Each of these four groups was further broken into two groups; the group who experienced academic success, and the group who met the antithesis of academic success (academic withdrawal). These groups were composed of the following individuals: College Success Withdrawal Total Arts and Sciences 752 149 901 Business Administration 387 88 475 Education 570 39 609 Journalism 247 27 274 Total 1,956 303 2,259 The BMD Biomedical Computer program was able to perform the stepwise discriminant analysis for each of the colleges in one operation. Therefore, all datawere loaded,
PAGE 35
-27 -together with the appropriate control cards, and submitted to the University of Florida's IBM 360 computer for analysis by the above described program. Summary The researcher used records of 2,259 students for the study sample. The nine variables included in the study were taken from the data bank maintained by the University of Florida and no new data were created for this study. Therefore, the study could be repeated for any group of students an investigator might select. The analysis tool, the BMD07M Stepwise Discriminant Analysis program, is an established, economical, analysis program that is widely available to investigators.
PAGE 36
CHAPTER III DEVELOPMENT OF THE MODEL Introduction The data collected for this study were submitted, in four parts, to the University of Florida's IBM 360 computer and subjected to discriminant function analysis using the BMD07M computer program. The colleges included in the study, namely Arts and Sciences, Business Administration, Education, and Journalism, made up these four parts. This portion of the report will consider the results of the analysis described above and examine each college individually together with the sets of mathematical models (the predictive model for academic success) that evolved from this analysis. Developing the Model The sample data for the four colleges included the records of 901 students in Arts and Sciences, 475 students in Business Administration, 609 students in Education, and 274 students in Journalism. The analysis of the data by the program led to the construction of a set of mathematical models for -2 8
PAGE 37
29 -each college ~one set of models for the prediction of academic success, and another for the prediction of academic withdrawal. The models consist of a constant (K) and a function for each of the variables (X) included in the study which when added together result in the prediction. The predictive model for each college consists of a set of two models, one for predicting success (S) and the other for predicting withdrawal (W). The models permit one to calculate figures for both academic su~cess a~d its antithesis, academic withdrawal, for any student for any of these colleges. This is possible because once weights have been established for the predictor variables by means of discriminant analysis the application of the weights to assign future cases to the groups tends to reproduce the composition of the original groups with respect to the measured variables. These calculated figures are the values for the discriminant function for the tested student and the smaller of the two figures is chosen as .the predicted result. That is, whichever is smaller, the (S) or the (W), is the prediction of success or withdrawal for that individual student.
PAGE 38
30 -Discriminant analysis seeks to make the predefined groups maximally different and to facilitate the assignment of new subjects to one of the given groups. If one merely wished the groups to have different means, one could find the means and multiply them by constants of whatever size one wished. To be meaningful, the difference in group means should be maximal relative to the score spread in each group. This is the reason for the program establishing the specific constants (K) as shown in these mathematical models for each college. If chance alone were applied to the data of this study, 100 per cent of the variability would be accounted for by chance. If one uses these models, 27 to 42 per cent of the variability is no longer accounted for by chance, but can be attributed to the variables included in this study. When one considers the overall college memberships in the data sample it will be noted that the models developed by the program correctly predicted academic success or academic withdrawal 72 to 79 per cent of the time for the individual colleges. This prediction was done on the 2,259 students included in the sample. The discrimination might be better described as the percentage of cases which are misclassified by the program of
PAGE 39
31 the total sample. These percentages ranged from 20 to 28 per cent for the four colleges. The efficiency of discrimination may be best expressed in correlational form. An appropriate question for this test might be: "What is the relationship (correlation) of the actual group membership (the groups as selected on the basis of the criterion variable) to the ability of the computer developed model to predict?" An appropriate correlation technique for this test is the use of phi(). Phi is a derivation of the Pearson r for dichotomous data and the data in this study may be considered dichotomous for statistical purposes. For a 2 x 2 table, a possible display of the results for each college of the predictive result (2) compared with (x) the actual result (2), is directly related to the chi-square (x2 ) statistic. For a discription of these techniques see Ostle (24), and Wyatt and Bridges (38). As a preliminary step in the development of phi, a chi-square coefficient of correlation was performed on the results of the analysis. This statistic was found to be significant at the .005 level of confidence for each of the college s in the study. This indicates a high degree of relationship between the predicted results based on the above predictive model for academic
PAGE 40
-32 -success and the observed results based on the criterion variable. In order to determine the amount of variance accounted for with this predictive model it was necessary to calculate phi (maximum) for each of the colleges. This was done so that the ratio of phi to phi(max) (/max) might be calculated. In order to assess the percentage of variance accounted for one must square the / max ratio. This value indicates that this percentage of the variance was accounted for by th~ variables included in the predictive model. These values ranged from 27 to 42 per cent in this study. Any correlation coefficient when squared may be interpreted as a percentage of variance accounted for. For a discussion of variance see Kerlinger (20), Ostle (24), or Wyatt and Bridges (38). Arts and Sciences The sample data for the College of Arts and Sciences included 901 students. By the criterion of academic success, described above, 752 students experienced success and 149 students experienced withdrawal, or the antithesis of academic success. The analysis of the data collected on each of these students led to the construction of a set of m athem atica l models:
PAGE 41
33 -s = -115.58421(K) + + 0.15270(X3 ) + + 9.63046(X&) + 0.50514(X9 ) w = -111.83672 (K) + + 0.16029(X3 ) + + 9.7515 0 ex 6) + 0.16460(Xs) where: S = Academic success 0.09268(X1) + 0.11371(X2 ) 5.37714(X4 ) + 8.58541(X5 ) 1.45148(X7 ) + 55.44241(X8 ) 0.07800(X1) + 0.10488(X2 ) 6.31623(X4) + 9.5100l(X5 ) 1.00025(X1) + 54.22322(X8 ) W = Academic withdrawal K = A constant X1 = Lower division grade point average X2 = Florida Twelfth Grade Test score or equivalent X3 = Age of student at beginning of junior year X4 = Sex of student X5 = Native or transfer student X6 = Florida or non-Florida resident X7 = Local residence X8 = Race of student X9 = Marital status of student at beginning of junior year The model for success for this college predicted correctly 77.5 per cent of the time, and the model for withdrawal predicted correctly 69.5 per cent of the time.
PAGE 42
34 When one considers the overall college membership in the sample this set of models predicted correctly 76.2 per cent of the time. The stepwise discriminant analysis loaded each variable for this college in descending order of whichever one would give the greatest improvement in classification. That is, the most significant variables for classification were loaded first. These variables were loaded in the following order: After all the variables had been entered, the cumulative approximate F ratio was found to be 26.51431 with 9 and 891 degrees of freedom. This is statistically a significant F beyond the .01 level, indicating little chance of these variables not being significant in the predictive model. A chi-square coefficient of correlation test was performed on the results of the analysis as a preliminary step in the development of phi. This was found to be: x2 = 131.8840, which was found to be significant at the .005 level of confidence. The phi coefficient was found to be: = 0.3825. This indicates a high degree of relationship between the predicted results based on the above
PAGE 43
-35 -predictive model for academic success and the observed results based on the criterion variable. In order to determine the percentage of variance accounted for by this predictive model, phi (maximum) was calculated. This was found to be: = 0.6749, and ~max the ratio of phi to phi was calculated to be: max ~max = 0.5668. This correlation coefficient when squared may be interpreted as a percentage of variance accounted for. This value indicates that 32.12 per cent of the variance was accounted for by this predictive model for academic success. Student Profile An analysis of the means of the variables included permits one to generalize the profile of these groups of students in the College of Arts and Sciences, those who succeed academically, and those who withdraw. Those students who succeeded had a higher lower division grade point average than those who witharaw. They also had a higher Twelfth Grade Test score than those who withdrew. Age was negative in its influence as the group who withdrew was older than the group who succeeded. The female student also withdrew from this college more often than did the male, as did the transfer student more than the native student. There was virtually no difference in success and
PAGE 44
36 withdrawal on the variable of residence or non-residence of the state of Florida. There was a difference in the two groups on the variable of local residence, indicating more successful students live in on campus married housing or off campus than in campus single housing or in a fraternity or soroity. The variable of race was virtually meaningless for this study as almost all students included in the study were white. Actually, more students refused to disclose their race on their university records than there were Negroes in the study. The variable of material status seems to indicate that the single student who had never married had the best chance for success in this college. Business Administration The sample data for the College of Business Administration included 475 students. Of this sample, 387 students met the criterion of academic success, described above, and 88 students met the criterion of academic withdrawal, the antithesis of academic success. The analysis of the data for this college by the BMD07M program led to the construction of a set of mathematical models for the prediction of academic success or academic withdrawal for this college:
PAGE 45
3 7 S = -169.13904(K) + 0.17927(X1 ) + 0.08873(X2 ) + 0.24582(X3 ) + 36.05937(X4 ) + 4.4952l(!s) + S.22244(Xs) + 3.0259S(X7 ) + 86.35260(X8 ) + 0.4676l(X9) W = -160.19824(K) + 0.16564(X1 ) + 0.08064(X2 ) + 0.24600(X3) + 35.90738(X4 ) + S.50448(X5 ) + S.4164l(X6 ) + 2.40330(X7 ) + 84.93285(X8 ) 0.79382(X9) where: S = Academic success W = Academic withdrawal K = A constant X1 = Lower division grade X2 = Florida Twelfth Grade point Test X3 = Age of student at beginning X4 = Sex of student Xs = Native or transfer student average score or equivalent of junior y ear Xs = Florida or non-Florida resident X1 = Local residence Xe = Race of student X9 = Marital status of student at beginning of junior year The model for success for this college predicte d correctly 74.2 per c ent of the time, and the mode l for withdrawa l predicte d correctly 68.2 per c ent o f the time.
PAGE 46
38 When the overall college membership in this sample is considered, this set of models predicted correctly 73.1 per cent of the time. The stepwise discriminant analysis loaded each variable for this college in descending order of whichever one would give the greatest improvement in classification. That is, the most significant variables for classification were loaded first. These variables were loaded in the following order: After all the variables had been entered, the cumulative approximate F ratio was found to be 11.83648 with 9 and 465 degrees of freedom. This is statistically a significant F beyond the .01 level, indicating little chance of these variables not being significant in the predictive model. A chi-square coefficient of correlation test was performed on the results of the analysis as a preliminary step in the development of phi. This was found to be: x2 = 57.544, which was found to be significant at the .005 level of confidence. The phi coefficient was found to be= 0.3480. This indicated a high degree of relationship between the predicted results based on the above
PAGE 47
-39 -predictive model for academic success and the observed results based on the criterion variable. In order to determine the per cent of variance accounted for by this predictive model, phi (maximum) was calculated. This was found to be: max= 0.66017, and the ratio of phi to phimax was calculated to be: /max= 0.52004. This correlation coefficient when squared may be inter-preted as a percentage of variance accounted for. This value indicates that 27.04 per cent of the variance was accounted for by this predictive model for academic success. Student Profile An analysis of the means of the variables included 1n the study permits one to generalize the profile of these two groups of students in the College of Business Administration, those who succeed academically, and those who withdraw. Those students who succeeded had a higher lower division grade point average than those who withdrew. They also had a higher Twelfth Grade Test score than the group who withdrew The group who withdrew was older than the group who succeeded, so age appears to be a negative influence for success in this college. The male student succee d e d more often than did the female in this college but the r e was very little difference in the means
PAGE 48
40 of these two groups. The native student succeeded more often than did the transfer student. There was virtually no difference in the means of the two groups with respect to the Florida or non-Florida resident variable, although more who withdrew were non-Florida residents. With respect to local residence, the means indicated that the student who succeeded lived in campus married housing or off campus, rather than in campus single housing or in a fraternity or sorority. The variable of race was virtually meaningless for ~his college in that almost all students included in the sample were white. The variable of marital status seems to indicate that the single student who had never married was more likely to withdraw than the married, or widowed, or divorced student. Education The sample data for the College of Education included 609 students. By the criterion of academic success, described above, 570 students experienced success and 39 students experienced withdrawal, or the antithesis of academic success. The analysis of the data collected on each of these students by the BMD07M program led to the construction of a set of mathematical models for the prediction of academic success or academic withdrawal for this college:
PAGE 49
41 -s = -106.61400(K) + 0 .10 4 21 (Xi) + 0.12112(X3 ) + 8.3598l(Xi+) + 17.29904(XG) + 0.91885(X7) 5.84442(X9) w --99.00226(K) + 0.09207(X1 ) + 0.11396(X3 ) + 9.17078(X4 ) + 16.98534(XG) + 0.53092(X7) 5.02412(X9) where: s = Academic success W = Academic withdrawal K = A constant + 0.06344(X2 ) + 8.7117S(Xs) + 42.93976(X8 ) + 0. 05771 (X2) + 9.10707(Xs) + 41.84727(Xa) X1 = Lower division grade point average X2 = Florida Twelfth Grade Test score or equivalent X3 = Age of student at beginning of junior year xi+ = Sex of student Xs = Native or transfer student x6 = Florida or non-Florida resident X7 = Local residence X8 = Race of student X9 = Marital status of student at beginning of junior year The model for success for this college predicted correctly 71.9 per cent of the time, and the model for withdrawal predicted correctly 71.9 per cent of the time.
PAGE 50
42 When the overall college membership in this sample is considered, this set of models predicted correctly 71.9 per cent of the time. The stepwise discriminant analysis loaded each variable for this college in descending order of whichever one would give the greatest improvement in classification. That is, the most significant variables for classification were loaded first. These variables were loaded in the following order: After all the variables had been entered, the cumulative approximate F ratio was found to be 4.39773 with 9 and 599 degrees of freedom. This is statistically a significant F beyond the .01 level, indicating little chance of these variables not being significant in the predictive model. A chi-square coefficient of correlation test was performed on the results of the analysis as a preliminary step in the development of phi. This was found to be: x2 = 32.7017, which was found to be significant at the .005 level of confidence. The phi coefficient was found to be: = 0.2317. This indicated a high degree of rela. tionship between the predicted results based on the above
PAGE 51
43 predictive model for academic success and the observed results based on the criterion variable. In order to determine the per cent of variance accounted for by this predictive model, phi (maximum) was calculated. This was found to be: = 0.39140, and the max ratio of phi to phimax was calculated to be: /max= 0.59197. This correlation coefficient when squared may be interpreted as a percentage of variance accounted for. This value indicates that 35.04 per cent of the variance was accounted for by this predictive model for academic success. Student Profile An analysis of the means of the variables included in the sample of the students of the College of Education permits one to generalize a profile of these two groups of students, those who succeed academically, and those who withdraw. Those students who succeeded had a higher lower division grade point average and higher Florida Twelfth Grade Test score than those students who withdrew. They were also older than the group who withdrew. The female student withdrew more often than did the male student. The native student succeeded more often than did the transfer student, as did the Florida resident more than the non-Florida resident. The means
PAGE 52
44 -of the variable of local residence indicate that the successful student lives in campus married housing or off campus, rather than in campus single student housing or in a fraternity or sorority. The variable of race was meaningless for this college due to the small number of Negro students in the sample. The variable of marital status seems to indicate that the single student who had never married had the best chance for success in this college, but there was very little difference for the other classifications of this variable. Journalism The sample data for the College of Journalism included 274 students. Of this college sample, 247 students met the criterion of academic success, described above, and 27 students met the criterion of academic withdrawal, the antithesis of academic success. The analysis of the data for this college by the BMD07M program led to the construction of a set of mathematical models for the prediction of academic success or academic withdrawal for students of the College of Journalism:
PAGE 53
45 -S = -125.20964(K) + 0.22712(X1) + 0.09616(X2 ) + 0.14281(X3) + 3.88128(X4 ) + 7.29087(X5 ) + 7.99617(X6 ) + 3.99100(X7 ) + SS.19489(X8 ) + 6.86018(X9 ) W = -121.91797(K) 0.20694(X1 ) 0.08989(X2 ) where: 0.15107(X3 ) -4.84507(X4 ) 9.16011(Xs) 8.0088S(X6 ) 3.36210(X1) -54.47794(Xa) S.522SS(X9) S = Academic success W = Academic withdrawal K = A constant X1 = Lower division grade point average X2 = Florida Twelfth Grade Test score or equivalent X3 = Age of student at beginning of junior year X4 = Sex of student X5 = Native or transfer student X6 = Florida or non-Florida resident X7 = Local residence X8 = Race of student X9 = Marital status of student at beginning of junior year The model for success for this college predicted correctly 79.8 per cent of the time, and the model for withdrawal predicted correctly 74.1 per cent of the time.
PAGE 54
-46 -When the overall college membership in this sample is considered, this set of models predicted correctly 79.2 per cent of the time. The stepwise discriminant analysis loaded each variable for this college in descending order of whichever one would give the greatest improvement in classification. That is, the most significant variables for classification were loaded first. These variables were loaded in the following order: After all the variables had been entered, the cumulative approximate F ratio was found to be 6.41046 with 9 and 264 degrees of freedom. This is statistically a significant F beyond the .01 level, indicating little chance of these variables not being significant in the predictive model. A chi-square coefficient of correlation test was performed on the results of the analysis as a preliminary step in the development of phi. This was found to be: x2 = 37.0809, which was found to be significant at the .005 level of confidence. The phi coefficient was found to be: = 0.36787. This indicates a high degre~ of relationship between the predicted results based on the
PAGE 55
47 above predictive model for academic success and the observed results based on the criterion variable. In order to determine the percentage of variance accounted for by this predictive model, phi (maximum) was calculated. This was found to be: 0.56435, and max the ratio of phi to phi was calculated to be max ~max .65190. This correlation coefficient when squared may be interpreted as a percentage of variance accounted for. This value indicates that 42.50 per cent of the variance was accounted for by this predictive model for academic success. Student ProfiZe An analysis of the means of the variables included in the sample for the College of Journalism permits one to generalize the profile of these two groups of students in this college, the group who succeed academically and the group who withdraw. Those who succeeded had a higher lower division grade point average, higher Florida Twelfth Grade Test score, and were younger than the group who withdrew. The male student was more successful than the female student. The native student was more successful than the transfer student, and the nonFlorida resident student was more successful than the Florida resident student. More successful students in
PAGE 56
48 -this college lived in campus married housing or off campus than lived in on campus single housing or in a fraternity or sorority. Race was a meaningless variable due to the few Negro students enrolled. The variable of marital status seems to indicate the successful student is married. Demonstration of the Predictive Model For a demonstration of how the predictive model for academic success wil work, three student data cards (identified as A, Band C) were selected from the data deck and processed for each of the four colleges. The students and their coded variables are as follows: X1 X2 X3 X4 Xs x6 X1 Xe X9 Student A 219 310 241 2 1 4 4 1 1 Student B 210 376 259 2 2 4 1 1 1 Student C 336 340 308 2 1 4 4 1 2
PAGE 57
49 See the sets of mathematical models above and Appendix V (for data coding) for complete identification of the coded variables and the procedure for arriving at the predictions. For student A the sets of mathematical models indicate that one should have expected little chance for academic success in either of the Colleges of Arts and Sciences or Journalism. The predictive model also indicates that this student would have been unlikely to achieve academic success in the College of Business Administration, but the prediction (indicated by the size of the difference between the discriminant functions) is not so strong as for the other two colleges. The model does predict that this student would have been likely to achieve academic success in the College of Education. The observed result for this student was that she did not achieve academic success as a student in the College of Arts and Sciences. For student B the predictive model indicate s that one could have expected the achievement of academic success in any of the four colleges included in the study. It does indicate tha t the chance for success in the College of Journalism should have b een expecte d to b e less tha n in any of the othe r three colle g es. The observe d
PAGE 58
so result for this student was that she did not achieve academic success as a student in the College of Journalism. For student C the calculations from the predictive model indicate that one could have expected this student to have not succeeded academically in any of the four colleges in the study. It does indicate that the most likely chance for academic success would have been in either the College of Arts and Sciences or the College of Education as the prediction for academic withdrawal for these two colleges was not strong. The observed result for this student was that she did achieve academic success as a student in the College of Arts and Sciences. Summary The researche r presented an analysis of the data resulting from the s tudy and the development of the pre dictive mode l for academic success in this portion of the report. Each college was treated separately with an analysis of the data and development of specific sets of mathematical models. When one considers the overall college memberships in the data sample it wi. 11 be noted that the model correctly predicte d academic success or academic withdrawal 72 to 79 p e r cent of the time for the individual colleges. Th e p e rcentage of misc l assif i e d c a s e s ranged from 20 to 2 8 p e r cent for the four colle g es.
PAGE 59
51 In order to determine the amount of variance accounted for with this predictive model a derivation of the Pearson r for dichotomous data (phi) was calculated for each college. This value indicates the percentage of variance accounted for ranged from 27 to 42 per cent for the four colleges. In three of the four colleges studied, Arts and Sciences, Business Administration, and Education, the Florida Twelfth Grade Test score had the heaviest loading. In the fourth college, Journalism, the native or transfer variable received the heaviest loading. The resident of Florida variable was generally the lowest factor in the loading. Generally speaking, the model will allow one to predict academic success, based on these variables, to a sufficient degree that they could be useful to those who counsel students in their academic lives. The possibility of failure by the mode l was demonstrated by the prediction made on randomly selected student C. The demonstration also showed that there are differences in colleges within the university and the selection of a college by a student has a significant influence on the probability of achieving academic success by a student.
PAGE 60
CHAPTER IV IMPLICATIONS AND RECOMMENDATIONS Introduction This section includes a discussion of theoretical implications developed from the study. It considers weaknesses of the predictive model for academic success and makes recommendations for its use. Recommendations are also made for further study. Implications The strongest point to be made from this study is that the probability of a student achieving academic success at the University of Florida may be directly relate d to his \choice of college enrollment within the university. Previous research seems to have been concerned with the overall population of a university and has not looked at the individual colleges within the overall structure. The sets of mathematical models for the four colleges included in this study point out that differences exist between colleges within the university. New colleges have used different admission requirements. The demonstration of the model shown above makes this point qu ite clear with Student A. It would 52 -
PAGE 61
-53 -. appear that this student might have achieved academic success had she been counseled to enter the college suggested by the predictive model. People who counsel students in their academic lives need to be aware of such possible different outcomes within the institution. The best possible choice, helped to be made by the use of the model, might make the difference between academic success and withdrawal from the institution. Weaknesses of the Model The model, statistically, has a built-in weakness due to the use of one of the variables. The Florida Twelfth Grade Test score is the sum of five percentile scores attained by the student on each of five separate academic area tests. The justification for including this variable is that it is commonly used throughout the ~tate of Florida and particularly at the University of Florida. It is used to rank students and thereby serves as a major criterion upon which to base admission to the University. This is especially true at the freshman level where a score of 300 is considered to be the minimum desired score. The statistical use of such a score (the sum of percentile scores) is open to question. The only defense offered is that this is common, acceptable practice in the state of Florida for this test series. It
PAGE 62
-54 wotild have been better to use raw scores, or a standard score in their place. Previous researchers have found the variable of race to be a significant discriminator. This was cited in the review of related literature. There were too few Negroes in the sample to allow this variable to have meaning in this study. The reason for this is that apparently few Negroes were enrolled as students at the University of Florida during the time of the study. This variable may become much more significant as a predictor in the years ahead. Use of the Predictive M o del The predictive model for academic success should be used at the University of Florida by those persons who counsel students in the selection of a college for study. It should be made clear to students that differences do exist between colleges with respect to the individual student's probability for academic success. The model should also be used within the four colleges as a tool for identifying potentially unsuccessful students. Those identified students might then be aided by a counselor or other helping person to achieve academ i c success. Forewained, they should be able to prepar e the mselve s b etter to achieve aca d e mic ally. This aid
PAGE 63
55 by a counselor to the individual student might be a stronger factor in his achieving academic success than all of the variables combined. Further Research There is a need for further research in the prediction of academic success. Specifimlly, the whole question of non-intellective variables and their contribution to the probability of academic success should be explored. Such variables as "local residence" while the student attends college, "marital status" at the beginning of college, and any changes that occur in this status, might be very important for educators to consider and need much more research. There are indications that these variables are more important in some colleges than others, but more research should be done. A greater in-depth study of the individual colleges is needed. Such a study should look at additional variables such as composition of the population of the college with respect to social class background, socio-economic status, size of home town, part time employment of itudents, and others. This type of information is not available on student r ecords but the review of related literature indicate s that it might be valuable in the refinement of a predictive model for academic success.
PAGE 64
56 Summary Success predictions are but one of the many things to be considered during the process of exploration with a student while attempting to help him direct his academic life. The predictions should serve as tools in the exploration process. As was suggested earlier by Justman (19), predictions should be tentative. Professional knowledge, judgment, and experience cannot be replaced by a predictive model for academic success, but c~n only be aided.
PAGE 65
APPENDIXES
PAGE 66
APPENDIX A STUDENT REGULATIONS AT THE UNIVERSITY OF FLORIDA Certain portions of the 1967 edition of The University Record which pertain to probation, suspension, and exclusion for academic reasons are presented. PROBATION, SUSPENSION, AND EXCLUSION FOR ACADEMIC REASONS The University of Florida is responsible for providing the best possible education in an economical and efficient manner. In order to discharge this -responsibility, the University must require reasonable academic progress from its students. Continuation of students who have demonstrated a lack of the necessary ability, preparation, industry, or maturity to benefit reasonably from a program of university study is inconsistent with the University's responsibility as a tax supported institution. The University of Florida Senate has enacted regulations covering probation, suspension, and exclusion. These regulations are directed toward enforcing the academic standards of the University. The academic standards of the University require both the maintenance of grade point averages consistent with a reasonable chance of satisfactory completion of the University programs and reasonable conformance to the catalog description of the program of study in which the student is engaged. Any college of the University may specify additional academic standards and students are responsible for observing the regulations pertaining to such standards. (p. 127) 58 -
PAGE 67
59 PROBATION BECAUSE OF UNSATISFACTORY ACADEMIC ACHIEVEMENT The purpose of academic probation is to recognize formally the fact that a student is making unsatisfactory progress. The conditions of academic probation are intended to: (1) Relate to quality of achievement below standards required for ultimate graduation; (2) Recognize unsatisfactory work at an early date; (3) Be sufficiently significant to make clear to the student, his parents, and the administration, the shortcomings of the student's performance; (4) Provide occasion for counseling; (S) Give students whose ultimate success is doubtful further opportunity to demonstrate adequate performance. All undergraduate students: Any student who is eligible to return to the University after a suspension because of academic reasons (failure to receive passing grades in at least one-half of his work or having his load reduced to less than twelve hours), will be placed on scholarship probation for his next quarter. In addition to University probation, a student may be placed on probation by the College in which he is registered if he does not maintain normal academic progress in the program of study in which he is engaged. Upper division students: An upper division student not on probation who fails to maintain a "C" (2.0) average for all work attempted in any quarter will be placed on scholarship probation for his next quarter. (p. 128) CONTINUATION OF PROBATION Upper division students: An upper division student on scholarship probation who withdraws prior to the final date published in the
PAGE 68
60 -University Catalog will be continuted on probation if his average is above "C" (2.0) for all work attempte d while registered in his present upper division college REMOVAL OF PROBATION Upper d ivision students: Scholarship probation will be remov e d for any full-time upper division student who earns a grade point average of "C" (2.0) or higher during the quarter when he is on scholarship probation at the University of Florida. Removal of college probation: A student will be removed from college probation when it is deemed by his college that he is making normal academic progress in the program of study in which he is engaged. (p. 128) SUSPENSION The purpose of suspension from the University for academic reasons is to remove from the University those students who would not ultimately meet requirements for graduation if they continued at their current level. of progress. The conditions of academic suspensions are intended to: (1) Select students whose performance indicates that they will not fulfill the requirements for graduation; (2) Encourage students to leave the University as soon as a high probability of failure is evident. All underg r a du ate students: Al 1 undergraduate students (i.e., students classified other than 7) who do not receive passing grades (A,B,C,D) in at least one-half of the hours attempted in any quarter shall be suspended immediately from the University. If such a student was not on probation for academic reasons and had not been previously suspende d for
PAGE 69
61 academic reasons, the suspension shall be for one ~uarter. If the student was on probation for academic reasons or had been previously suspended for academic reasons, he shall be excluded from the University without the opportunity to re-enroll. However, failure in only one course carrying five quarter credits or less shall not cause the student to be suspended under this provision. All undergraduate students who are dropped from a course because of excessive absences or unsatisfactory work and as a result of such action are left with an academic load of less than twelve credits, shall be suspended immediately. If such a student was on probation or had been previously suspended for academic reasons, he shall be excluded from the University with out the opportunity to re-enroll. If the student was not on probation or had not been previously suspended f~r academic reasons, the suspension shall be for one full quarter. Any student who receives a second EW (dropped for non-attendance or unsatisfactory work) in military science courses will be suspended from the University for one full quarter. Upper division students: An upper division student who is on scholarship pTobation will be ineligible for further registration at the University unless he maintains an average of "C" (2.0) in all work attempted that quarter or has a:n average of "C" (2.0) in all work attempted while registered in his present upper division college. (p. 129) EXCLUSION Exelusion from an undergraduate program of study: A student may be excluded from a program of study by the College responsible for the program if he fails o r refuses to maintain normal academic progress. Such exclusion does not prohibit the student from enrolling in other programs or colleges if he meets the requirements. (p. 130)
PAGE 70
62 -The student regulations were changed in the years between 1967 and 1970. For that reason certain portions of the 1970 edition of The University Record which pertain to probation, suspension, and exclusion for academic reasons are included. PROBATION BECAUSE OF UNSATISFACTORY ACADEMIC ACHIEVEMENT The purpose of academic probation is to recognize formally the fact that a student is making unsatisfactory progress. The conditions of academic probation are intended to: (I) Relate to quality of achievement below standards required for ultimate graduation; (2) Recognize unsatisfactory work at an early date; (3) Be sufficiently significant to make clear to the student, his parents, and the administration, the shortcomings of the studentts performance; (4) Provide occasion for counseling; (S) Give students whose ultimate success is doubtful further opportunity to demonstrate adequate performance. AZZ und~rgraduate students: A student with less than a 2.0 grade point average in his respective division (lower, upper) shall be placed on scholarship ~arning if he has a grade point deficit of nine or less. A student with less than a 2.0 grade point average in his respective division shall be placed on scholarship prolatio n if he has a grade point deficit of ten or more but less than twenty. Any student who is eligible to return to the University after a suspension because of academic reasons will be placed on scholarship probation for his next quarter. In addition to University probation> a student may be placed on probation by the College in which he is
PAGE 71
63 registered if he does not maintain normal academic progress in the program of study in which he is engaged. (p. 161) CONTINUATION OF PROBATION All undergraduate students: A student's scholarship warning shall be continued as long as he has a grade deficit of one but not more than nine. A student's scholarship probation shall be continued as long as he has a grade point deficit of ten but not more than nineteen. If his grade point deficit places him in another probation category, he shall be subject to the provisions of that category. REMOVAL OF PROBATION All undergraduate students: Scholarship probation and/or scholarship warning will be removed when a student's grade point deficit in his respective division has been reduced to zero. Removal of college probation: A student will be removed from college probation when it is deemed by his college that he is making satisfactory academic progress in the program of study in which he is engaged. SUSPENSION The purpose of suspension from the University for academic reasons is to remove from the University those students who would not ultimately meet requirements for graduation if they continued at their current level of progress. The conditions of academic suspensions are intended to: (1) Select students whose performance indicates that they will not fulfill the requirements for graduation; (2) Encourage students to leave the University as soon as a high probability of failure is evident.
PAGE 72
-64 -AZZ undergraduate students: A student with a grade point deficit of twenty or more in his respective division (lower, upper) shall be suspended from the University for one quarter. A student re-enrolling after a one quarter suspension will be on final scholarship probation, and if his grade point deficit is twenty or more at the end of the quarter he re-enrolls, he will be suspended without the possibility of re-registering except by committee action. (p. 162) EXCLUSION AZZ undergraduate students: A student may be .excluded from a program of study by the College responsible for the program if he fails or refuses to maintain normal academic progress. Such exclusion does not prohibit the student from enrolling in other programs or colleges if he meets the requirements. (p. 163) CLASSIFICATION OF STUDENTS~FLORIDA OR NON-FLORIDA For the purpose of assessing fees, applicants shall be classified as Florida or non-Florida students. A Florida student is a person who shall be a citizen of the United States or a resident alien and who shall have resided and had his habitation, domicile, home and permanent abode in the State of Florida for at least twelve (12) months immediately preceding his current registration. In applying this regulation, "applicant" shall mean a student applying for admission to the institution if he is married or 21 years of age, or, if he is a minor, it shall mean parents, parent, or legal guardian of his or her person. In all. applications for admission by students as citizens of the State, the applicant, if married or 21 years of age, or, if a minor, his parents or legal guardian shall make and file with such application a
PAGE 73
65 written statement under oath that such applicant is a bona fide citizen and resident of the State and entitled as such to admission upon the terms and conditions prescribed for citizens and residents of the State. In the determining of a Florida resident for purposes of assessing fees, the burden of proof is on the applicant. Under the law an applicant can change his place of residence from another state to the State of Florida only by actually and physically corning into the State and establishing his residence with the intention of permanently residing within the State. The domicile or legal residence of the wife is that of the husband, and the legal residence of a minor is that of the parents, parent, or legal guardian of his person. A non-Florida student may apply in writing for reclassification prior to any subsequent registration under the provisions set forth below. To qualify for reclassification as a Florida student, a person (or if a minor, his parents) shall have resided in Florida for twelve (12) months, shall have filed a declaration of intent to become a resident of the State, and shall be registered to vote in the State. An alien shall have resided in Florida for twelve (12) months and must present U. S. Immigration and Naturalization certification that he is a resident alien. If the application is supported by evidence satisfactory to the University that the student then qualifies as a Florida student, his classification will be changed for future registrations. (p. 139)
PAGE 74
CoZ. 02-10 11-12 13 14-16 17-19 20-22 23 24 25 26 27 28-30 31-33 34-35 36-41 APPENDIX B COMPUTER CARD DATA CODING Item Social Security Number College Membership 01 if AS, 02 if BA, 03 if ED, 04 if JM Academic Success 1 if graduated, blank if not Lower Division G.P.A. (Freshman and Sophomore years) 12th Grade Test Score or equivalent Age in Months from birth to entry into Junior year. Possible Junior starting times are 9/67, 1/68, 4/68, 6/68 Sex 1 if male, 2 if female Native or Transfer -1 if native, 2 if transfer (Native is less than 11 transfer hours) Reiident 4 if Florida, 1 if non-Florida, 3 if Alien Residence -1 if Campus, 2 if Frat. or Sor., 3 if Campus Married, 4 or 5 if Off Campus Rae~ -1 if White, 2 if Negro Number of times on probation Number of times student was part-time since becoming a Junior Number of term first pa.rt-time occurred (Fall 67==05 ) SCAT Scores 66 -
PAGE 75
42-47 48-51 52-53 54 55 56-57 58-60 61-62 63-64 65-80 67 -SAT Scores Matriculation Information (Term by month and year, status) Institution Code (From which matriculated) Marital Status as of first Junior status -1 if single, 2 if married, 3 if divorced, 4 if widowed Second marital status if different from above Number of term in which first change in marital status occurred Transfer Hours Terms enrolled while in Junior or Senior status Term student entered his Junior year of study (Fall 67=05, Winter 68=06. ) Name of Student
PAGE 76
BIBLIOGRAPHY 1. Anderson, Harry E., "Regression, Discriminant Analysis, and a Standard Notation for Basic Statistics," in Handbook of Multivariate Experimental Psychology. Edited by Raymond B. Cattell, Chicago: Rand McNally and Company, 1966. 2. Atwell, Charles Alan, "Institutional and Community Characteristics Related to the Effectiveness of Transfer Programs in Florida Public Junior Colleges,'' Doctoral Dissertation, University of Florida, 1968. 3. Bashaw, Wilbur Louis, "A Central Prediction System for Predicting the Success of Junior College Transfers in Florida Universities," Doctoral Dissertation, Florida State University, 1963 ( Dissertation Abstracts, Vol. 25, p. 282). 4. Bauer, Roger, William A. Mehrens and John F. Vinsonhalcr, "Predicting Performance in a Computer Programming Course," Educationa l and Psyc h olog ical Meas u rement, 1968, Vol. 28, No. 4, pp. 1159-1164. 5. Bayes, Andrew Hartin, "An Application of Hotelling's Canonical Correlation to Academic Prediction," Doctoral Dissertation, University of Miami, 1968 (Dissertation Abstra c ts, Vol. 29, p. 2512). 6. Bottenberg, Robert A., and Joe H. Ward, Jr., Applied Multiple Linear Regression. Technical Documentary Report PRL-TDR-63-6. United States Dep artment o f Commerce. Washington: U. S. Government Printing Office, 1963. 7. Byron, Anthony R., "Non-Intellective Variables Related to Successful and Unsuccessful Students in a Junior College/' University o f Missouri, 1968. ERIC Microfiche ED 023 387. 8. Clarke, Johnnie R., and Rose Mary Ammons, "Identification and Diagnosis of Disadvantaged Students,'' J unio r College Journal, 1970, Vol. 40, No. 5, pp. 13-17. 9. Conklin, R. C., a nd D. G. Ogston, "Predictio n of Acad endc Succe s s for Fres h men at the University o f Calgary," -68 -
PAGE 77
69 Atberta Journal of Educational Research, 1968,Vol. 14, No. 3, pp. 185-192. 10. Cooley, William W. and Paul R. Lohnes, Multivariate Procedures for the Behavioral Sciences. New York: John Wiley and Sons, Inc., 1962. 11. Cooper, Leland Ross, "The Relationship of Selected Factors to the Continuance of Junior College Graduates at Senior Institutions," Doctoral Dissertation, University of Florida, 1964. 12. Dixon, W. T., editor, BMD: Biomedical Computer Programs, Berkeley and Los Angeles: University of California Press, 1968. 13. Fishman, Joshua A., "Some Social-Psychological Theory for Selecting and Guiding College Students," The American College. Edited by Nevitt Sandford, New York: John Wiley and Sons, Inc., 1962, Part V, pp. 666-689. \ 14. Gadzella, Bernadette, and Grace Bentall, "Differences x 15. I 16. 17. 18. \ 19. in High School Academic Achievements and Mental Abilities of College Graduates and College Drop-Outs," College and University, 1967, Vol. 42, No. 3, pp. 351-356. Hopper, Harold H., Predictors of College Success, 1968, ERIC Microfiche ED 024 374. Iffert, Robert E., Retention and Withdrawal of College Students. United States Office of Education Bulletin No. 1. Washington: U. S. Government Printing Office, 1968. Irvine, Donald W., "Graduation and Withdrawal: An Eight-Year Follow-up," College and University, 1965, Vol. 41, No. 1, pp. 32-40. Ivanoff, John M., "The Use of Discriminant Analysis for Predicting Freshman Probationary Students at One Midwestern University," Educational and Psycho logical Measurement, 1961, Vol. 21, No. 4, pp. 975-985. Justman, Joseph, "The Counselor's Use of Measurement in Prediction," National Catholic Guidance Con ference Journal, 1968, Vol. 12, No. 2, pp. 145-153.
PAGE 78
70 20. Kerlinger, Fred N., Foundations of Behavioral Research. New York: Holt, Rinehart and Winston, Inc., 1964. 21. Lavin, David E., The Prediction of Academic Performance. New York: Russell Sage Foundation, 1965. \ 2 2. Magoon, Thomas M. and Martha J. Maxwe 11, "Demographic Differences Between High and Low Achieving University Students," The Journal of College Student Personnel, 1965, Vol. 6, pp. 367-373. 23. Maxwell, A. E., "Canonical Variate Analysis When the Variables Are Dichotomous," Educational and Psy chological Measurement, 1961, Vol. 21, No. 2, pp. 259-271. 24. Ostle, Bernard, Statistics in Research. Ames: The Iowa State University Press, 1963. 25. Renetzky, Alvin, editor, Yearbook of Higher Education~ 1969. Los Angeles: Academic Media, Inc., 1969. '26. Richards, James M., Jr., John L. Holland, and Sandra W. Lutz, "Prediction of Student Accomplishment in College," Journal of Educational Psychology, 1967, Vol. 58, No. 6, pp. 343-355. 27. Rose, H. A., and C. F. Elton, "Another Look at the College Drop-out," Journal of Counseling Psychology, 1966, Vol. 13, pp. 242-245. \ 28. Roudabush, Glenn E., "A Study in Prediction from Biographical Information," Doctoral Dissertation, University of Washington, 1963 (Dissertation Abstracts, Vol. 25, p. 1324. 29. Schroeder, Wayne L., and George W. Sledge, "Factors Related to Academic Success," The Journal of College Student Personnel, 1966, Vol. 7, No. 2, pp. 97-104. 30. Sims, David M., "A Study of the Relationship of Selected Institutional Characteristics of the Junior College of Origin to the Academic Performance of Public Junior College Transfer Students in the University System of Florida," Doctoral Dissertation, Florida State University, 1966. ERIC Microfiche ED 026 241.
PAGE 79
-71 -31. Summerskill, John, "Dropouts from College," The American College, Edited by Nevitt Sandford, New Y~rk: John Wiley and Sons, Inc., 1962, Part V, pp. 627-657. 32. The University Record, Undergraduate Catalog Issue, Gainesville: University of Florida, 1967. 33. The University Record, Undergraduate Catalog Issue, Gainesville: University of Florida, 1970. 34. Thornton, James W., Jr., The Community Junior College, 2nd Edition. New York: John Wiley and Sons, Inc., 1966. 35. Trent, James W., and Leland L. Medsker, Beyond High Schoo Z. San Francisco: Joss ey-Bas s, Inc. Pub-1 ishers, 1968. 36. Walker, John E., Aqademic Performance of Native and Transfer Students in the Upper Division of the University of Florida, 1966-1968. Gainesville: Institute of Higher Education, 1969. 37. Woodward, Ivor Carey, "The Relative Efficiency of Multiple Regression Analysis and Multiple Cutoff Analysis in the Prediction of Academic Performance in a Selected Medical School," Doctoral Dissertation, University of Southern California, 1968 (Dissertation Abstracts, Vol. 29, p. 1143). 38. Wyatt, Woodrow, and Charles M. Bridges, Jr., Statistics for the Behavioral Sciences. Boston: D. C. Heath and Company, 1966.
PAGE 80
BIOGRAPHICAL SKETCH Ronald Wheeler Mcfaddin was born August 31, 1934, at Cincinnati, Ohio. He attended elementary school in San Diego, California, and graduated from Manatee County High School in Bradenton, Florida, in 1952. He attended the University of Florida from 1952 to 1954 when he entered the Air Force and served as a pilot and administrative officer. He was released from military service in 1958 and took employment in iales and management. Returning to the University of Florida in 1965 he resumed studies and received the Bachelor of Science degree in Health Education in 1967. He enrolled in the Graduate School of the University of Florida where he studied counseling and guidance. He received the Master of Education degree in 1968 with a major in Personnel Services, and since that date has been studying toward the Doctor of Education degree with a major in Educational Administration with emphasis on higher education. Ronald Wheeler Mcfaddin is married to the former Mary Katharine Moore and is the father of two children, David and Laurie. He is a member of Phi Delta Kappa and Kappa Delta Pi. 72 -
PAGE 81
I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Education. es L. Wattenbarger, ofessor of Education I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Education. I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the d:gree of D::t:~n. -Professor of Education I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the de Docto f Education.
PAGE 82
I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Education. ,,,----/) // // ,:',,/J \ ----lt~., i I, i/0~""Y Jofo?Iy. James /// As-s'o ciate Professo, f of Management / / : i ./"/ This dissertation was submitted to the Dean of the College of Education and to the Graduate Council, and was accepted as partial fulfillment of the requirements for the degree of Doctor of Education. March 1971 D~ E
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EJOYU8S8P_3V71L1 INGEST_TIME 2017-05-22T20:23:18Z PACKAGE AA00054895_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
|
|