MapWorks Data and Retention Rates - Part I

This paper details a statistical project requested by St. Cloud State University to predict Term 2 and Term 3 enrollment, also known as retention. Finding out what affects retention can be important because it helps faculty and staff determine the most at-risk students. If students can be found and helped, then the university can prevent them from dropping out altogether. Some things, like grades or money, are obvious influences on retention. However, there tend to be students with financial means and high GPA that will suddenly drop out. Why do these promising students decide to leave? So, the main focus of this project is to identify underlying factors that may be affecting retention. Most of the data that this project is based on comes from the fall of 2015. It was collected through the MapWorks ‘Transitions’ survey, which is a survey given in September to college students to analyze their performance and retention in their first year of college. The 218 survey questions were divided into 22 factor groups according to the topics addressed. This data was used to create models. Two indexes where created from a selection of the Factors: those determined to be indicators of academic competency, and those determined to measure a sense of belonging on the campus. In this case, academic competency involves aspects like talking to professors and getting homework completed. A sense of belonging is how integrated and comfortable a student is on campus. The indexes were created so that the effects of less obvious factors could be measured. The index created from the Factors that indicated academic competency was called the Academic Index. The other index, created from the belonging-related factors, was called the Belonging Index. Data analysis was done by creating models in JMP Pro 13. The variables being predicted were Term 1 GPA, Term 2 Enrollment, and Term 3 Enrollment. The indexes were tested first against the GPA from the first term. Since GPA is an already known indicator of retention, it was a good mid-year variable to see the effect of the indexes. The Academic Index turned out to be the better indicator of mid-year performance than the Belonging Index. After some more modeling was done to test the need for any interaction terms, final models were made to predict enrollment for Term 2 and Term 3. The variables QPP, an indicator of academic success prior to college; and Pell Eligible Flag, a measure of a student’s finances; were included along with the two indexes. Results show that the QPP and Belonging Index variables were very good predictors of both Term 2 and Term 3 retention. The other two variables, Academic Index and Pell Eligible, had less important effects in the model. So, in future projects the belonging variable could help produce better predictions of retention