The Repository @ St. Cloud State

Open Access Knowledge and Scholarship



Torrence Savage

Document Type

Research Study

Publication Date

Spring 2018


The St. Cloud State University Belonging Team, under the direction of Dr. David Robinson, has been investigating how students’ sense of belonging affects their retention at the university. It was found in previous Belonging Team research that Belonging Index, Term 1 Cumulative GPA, Term 1 Credit Completion Rate and other Demographic variables were useful in prediction.

The goal for this project was to create two models for predicting the likelihood of Term 3 retention after a student’s first semester. These models are based on the Fall 2014 and 2015 cohorts of new first-year domestic students. The data includes measures of academic success, referred to as Academic Outcomes, including GPA, credits attempted, and credits completed among others. The demographic information includes variables of interest including gender, whether or not SCSU is the closest university to the student’s home, financial aid gap, etc. Students’ sense of belonging was measured by a series of questions in an online survey designed by the St. Cloud State University Belonging Team. There were 10 questions asking about topics such as a student’s commitment to complete their degree and how much they regretted leaving home. The answers to these questions were measured on a 7-point scale which was converted to a new 5-point scale variable called Belonging Index. This was done to make the data comparable to data collected in later years with a survey using a 5-point scale.

The first model created has predictors that include only Belonging Index and Academic Outcomes. The second model has Belonging Index, Academic Outcomes, and Demographic variables as predictors. These two models are compared with a previous model, created in Fall 2017 by the author, which only included Belonging Index and Demographics as predictors. A final model, shown in Table 6, was developed based on model accuracy, statistical and practical significance of the predictors, and the complexity of the model. The model was applied to the Fall 2017 cohort of new first-year domestic students to find their average predicted probability of Term 3 enrollment and compare it with the Fall 2014-2015 cohorts.

It was found that a student’s original Belonging Index is still a very strong predictor of Term 3 enrollment after a student’s first semester. Also, Term 1 Cumulative GPA and Term 1 Credit Completion Rate were the best predictors of Term 3 enrollment among those tested.

The average predicted probability of Term 3 enrollment for the Fall 2017 cohort, after Term 1 and excluding Term 2 dropouts, was 79.9% compared to an average Term 3 enrollment of 79.1% for the 2014-2015 cohorts.

The predicted probabilities of enrollment for all students of the 2017 cohort were saved and sent to Dr. Robinson. They could then be grouped according to predicted probability of enrollment, and students could be targeted for help based on categories of student most at risk.


Edited by Shaya Kraut

Creative Commons License

Creative Commons Attribution-NonCommercial 4.0 International License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License