122 - Using Binary Logistic Regression to Predict Graduation Status and Enhance an Early Warning Response System

Abstract Information: The purpose of this poster is to share how graduation prediction models enhanced an internal Early Warning Response System (EWRS) in a K-12 setting. Specifically, the audience will learn about graduation prediction models, see the functionality of a current EWRS, and learn how evaluators can apply a new indicator to help predict on-time graduation. The goal of this research was to utilize predictive modeling on school-level data to assess outcomes for graduating on-time in a large S.C. school district. Practical goals were to embed predictor variables into a data dashboard which serves as a district-wide, internal EWRS. The results of this research informed district administration in middle and high schools about the particular outcomes relevant to graduation likeliness and how statistics may help identify students at-risk for not graduating on-time. Interestingly, a new indicator was developed called a Diploma Indicator, which flags students if they have not received a certain combination of classes and credits by the end of their 9th grade year. The statistically significant predictor variables for the middle school model was 6th grade attendance, number of 6th grade course failures, overage, and 5th grade Spring Math and Reading MAP scores. For high school, credit obtainment in math and English by the end of the 9th grade year and end-of-course exams (in particular math and English) proved critical for on-time graduation at the high school level. The limitations to this work centered around the data available to researchers and the model’s inability to predict non-graduates well. Conversations about grade floor and extensive content recovery policies helped explain some of the variation seen in non-graduates and those conversations should continue. New and veteran evaluators will learn about graduating prediction models and how these results were the impetus for a new standard in an internal EWRS. The overall goal of this work was to better understand how to flag students who were off-track to graduate much earlier so interventions could be implemented before disengagement or dropout occurs.