In this talk, I describe the development and deployment of the DPI Dropout Early Warning System predictive model for student risk of dropout. This example serves as a backdrop to explore the issues and approaches of applied predictive modeling, their differences from inferential modeling, and the tradeoffs inherent in deploying a predictive model within a large organization. Time will be taken to also discuss the major R components of the project and the features specific to R that enabled the project to move forward.
Jared Knowles currently serves as a research analyst with the Wisconsin Department of Public Instruction. There he has led the design and deployment of the Wisconsin Dropout Early Warning System (DEWS) and has worked on numerous policy analyses for the department. He focuses on ways to display these results in ways that are interpretable and actionable by decision makers. He is currently completing his PhD in political science at the University of Wisconsin-Madison. He is also a fellow in the Interdisciplinary Training Program in Education Sciences, an IES pre-doctoral training program in the Wisconsin Center for Education Research