STATISTICAL MODELS to IDENTIFY STUDENT CHARACTERISTICS to TAILOR GRADUATION INITIATIVES

• Activity Number: 189 - Contributed Poster Presentations: Section on Statistical Education
• Type: Contributed
• Date/Time: Monday, July 29, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and Data Science Education
• Author(s): Ayona Chatterjee* and Chinki Rai and Fanny Yeung
• Companies: California State University East Bay and CSUEB and CSUEB
• Keywords: Graduation Rates; Logistic Regression; CART; Confusion Matrix; At risk students

Abstract

Improving graduation rates is one of the biggest missions in many universities across the country. The work here presents a statistical tool box to use early academic performance as a predictor for graduation with logistic regression, classification and regression trees (CART) with machine learning techniques. Our study utilized data from one academic cohort across 6 years to identify significant student academic characteristics that are related to graduation. Various statistical methods such as logistic regression, survival analysis and neutral networks have been proposed to study graduation rates. In this study, we present results from both a logistics regression and CART analysis and contrast and compare results. Further we create training and test data sets to check for validity of our model on a different data set. The model can then be applied to current students finishing their freshmen year and assign probabilities to successfully graduate in a pre-determined framework. The study and the significant factors are specific to the institutions’ campus but the model allows the study to be replicated on any campus to support graduation initiatives.