Senior Honors Thesis: "Semiparametric Modeling of Train Accident Costs with Implications for the Importance of Positive Train Control"

Abstract:  We study the factors influencing the total cost of train accidents of various types for the years 2009-2015, using data from the Federal Railroad Administration Office of Safety Analysis. The goal of such an analysis is to develop a good predictive model for accident costs in future years. This can provide some evidence regarding the potential efficacy of the Positive Train Control (PTC) system, which is currently in development in the United States. The analysis is conducted using generalized additive models (GAM) fitted by penalized iteratively reweighted least squares (P-IRLS). Testing procedures for such models are described and implemented, and robustness checks are performed prior to a final interpretation of the results.

Advisor: Todd Kuffner