BEHAVIORAL, CURRICULAR AND DEMOGRAPHIC VARIABLES AFFECTING UNIVERSITY STUDENTS' CHANGE IN DEGREE MAJOR PROGRAMS (FLORIDA; LOG-LINEAR MODELS, INSTITUTIONAL RESEARCH)

The purpose of this study was to examine the proposition that a change in major by a college student is determined in part by vocational preference and in part by the curricular constraints within a given college or university. The study used both the vocational preference groups proposed by Holland (1966, 1973) and curricular constraints proposed by Lightfield and Rice (1973; Rice, 1973).;The principal activities of the study consisted of describing the major changes made by the population of students who changed majors, fitting a log-linear model to a multiway contingency table, and cross-validating the model on a cross-validation group.;The study was designed to determine whether changes in academic major follow the order of Holland's typology, and whether these changes are limited by curricular constraints and can be predicted from the initial major assigned and the curricular constraints proposed by Rice. The results of the study supported the proposition that the changes follow the order of Holland's typology. They did not support the proposition that changes in academic major are limited by curricular constraints and that those changes can be predicted from the initial major using the curricular constraints proposed by Rice.;Two variables--age and sex--were determined to be related to curriculum and were added to the data, forming a five-way contingency table. A model was successfully fit to the data of the five-way contingency table using .80 adequate R('2) as the criterion for fit.;The data for 15 academic quarters were collected from 36,611 student records from the nine universities of the State University System of Florida. This population was sorted into a model construction group (N = 18,309) and a cross-validation group (N = 18,302).;The five variables--vocational preference, type of major change, the catalytic action of restrictions, sex, and age--were successfully fit into a model which explained almost all the student behavior captured in existing university records.