B3H: Integrating Poisson regression into the undergraduate curriculum

Abstract

Students who know multiple linear regression techniques have developed a strong foundation from which they can build many related, useful models. Poisson regression is one such model, appropriate when the response variable is a count (e.g. residents per household or goals per game). In this session, we will discuss strategies for transitioning from multiple linear regression to Poisson regression through several real case studies. We will compare and contrast linear and Poisson regression models, and together we will discuss potential modifications to the Poisson regression model in cases where core assumptions are violated (e.g. too many zeros or correlated observations). In our experience, students feel greatly empowered when they build on their linear regression foundation to model real case studies that they previously could not handle because of violations of core linear regression conditions. Ultimately, participants in this session will have an opportunity to consider where Poisson regression might fit and be seamlessly scaffolded in their curriculum, although this session will also offer ideas to participants who are simply curious about Poisson regression and its applications. Participants will ideally have worked with multiple linear regression models before, and then through this session they will gain a repository of case studies, active learning lesson plans, and open source online materials (including the book Beyond Multiple Linear Regression) for use with their students or in their own research.