Attrition IDENTIFIERS *Logit Analysis

Studies of 1-year retention commonly use a binary outcome (retained after 1 year, yes/no) as the dependent variable. This paper examines the use of alternate specifications of 1-year retention that include information about spring semester stopout behavior. Different types of discrete choice models are first described, Next, the multinomial logit model using four outcomes (enrolled first fall semester only; enrolled first fall semester and spring semester; enrolled first fall semester, stopped out and returned the following fall; and enrolled all three semesters) is compared with the traditional dichotomous logit approach using the 3,589 first-time full-degree seeking freshmen enrolled at a large university. The pseudo R-square indicates the multinomial model fits the data better, but for practical reasons it has very poor predictive power. The multinomial model does yield interesting results for the impact of independent variables on retention; for example, increases in high school grade point average impacts the probability of a student being in the fall-spring outcome more than the probability of falling in the first fall semester only outcome. (Contains 5 tables and 13 references.) (Author/SLD) Reproductions supplied by EDRS are the best that can be made from the original document. Viewing One-Year Retention as a Continuum: the Use of Dichotomous Logistic Regression