This paper describes a methodology used in an on-going retention study at New York University (NYU) to identify a series of easily measured factors affecting student departure decisions. Three logistic regression models for predicting student retention were developed, each containing data available at three distinct times during the first semester: first, prior to the start of the fall semester; second, after the third week of classes; and third, at the end of the first semester. A method of identifying appropriate variables for inclusion in the logistic regression model is discussed as well as a rationale for choosing different cut points to classify the logit results. The study followed Fall 1994 and Fall 1995 freshmen (n=2209) among whom 272 students did not return to NYU a year after entry. Variables were grouped into six general categories describing: (1) family background/individual attributes; (2) pre-college schooling; (3) institution commitment; (4) first-term academic integration; (5) first-term social integration; and (6) first-year finances. The study found all three models were reasonably effective in identifying high risk students using various probability cutoff points. It concluded that use of all three models to identify students at risk at the three different times would allow for an optimum intervention strategy. (BF)