This is the fifth in a series of articles describing an attrition prediction and intervention project at the University of South Florida (USF) in Tampa. The project was originally presented in the 83(2) issue (Miller 2007). The statistical model for predicting attrition was described in the 83(3) issue (Miller and Herreid 2008). The methods and approaches for intervening with students at highest risk of attrition were discussed in the 84(3) issue (Miller and Tyree 2009). A second model, based upon the students who entered in 2007, was described in the 84(4) issue (Herreid and Miller 2009).
The work described in previous articles addressed predicting risk of freshman attrition. The two logistic regression models, based on the classes that had entered USF in 2006 and 2007, predict attrition between the beginning of the freshman year and the beginning of the sophomore year. The data that contributed to the model were the results of the administration of the College Student Expectations Questionnaire (CSXQ) combined with demographic and academic data collected by the University. The project described in this article produced a model for predicting the risk of attrition of individual students between the beginning of the sophomore year and the beginning of the junior year.
EARLIER FRESHMAN TO SOPHOMORE PREDICTIVE MODELS
Factors in the data set collected from the students who entered in 2006 that distinguished between dropouts and persisters in the first year of college and that became part of the attrition prediction formula were:
* High school grade point average (with a positive effect on persistence)
* Being Black versus being White (positive)
* Expecting to participate in student clubs and organizations (positive)
* Expecting to read many textbooks or assigned books (positive)
* Expecting to read many non-assigned books (negative)
* Expecting to work o if campus (negative)
In the article in the 83(3) issue oí College and University (Miller and Herreid 2008) we discussed these variables in detail. The reader is reminded that the variables have predictive value, but their relationship to persistence is not necessarily causal. The researchers were able to conduct an early test of the model and of the intervention program, the Mentoring Project. The model is intended to predict attrition between the start of the fall semester of the first year and the start of the fall semester of the second, or sophomore, year. However, the researchers reviewed data regarding fall-to-spring attrition to determine if there were results that showed the value or power of the model. Those results, showing the attrition rate, fall-to-spring for the class beginning in the summer or fall of 2008:
These results seem to suggest that the model has some predictive value, and they also suggest that the Mentoring Program is having a positive effect in modifying risk. The Fall Semester will give a better understanding of the accuracy of the model, but the early signs are encouraging.
In 2007, the response rate from students was much better than the previous year, and the researchers had about 2,700 usable surveys with identifying information. The logistic regression application showed the following variables to have predictive merit, and they were included in the model:
* High school GPA (again, with a positive effect on predicting persistence)
* Being Asian vs. being White (positive)
* Being Black vs. being White (positive)
* Scoring higher on the SAT Combined measure (negative)
* Expecting to use library and internet resources (positive)
* Expecting to read many non-assigned books in college (negative)
* Being enthusiastic about college (positive)
* Believing that the university will emphasize developing aesthetic, expressive and creative qualities (negative)
* Expecting to work o if campus while in college (negative)
CONCERN FOR AuRITION AFTER SOPHOMORE YEAR
The sophomore year experience is the subject of some concern and has attracted attention at institutions of higher education. …