Causal Inference for the Risk of Cellphone Use While Driving

• Activity Number: 415 - Modeling in Transportation Safety Issues
• Type: Contributed
• Date/Time: Tuesday, July 31, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Transportation Statistics Interest Group
• Author(s): Danni Lu* and Feng Guo and Fan Li
• Companies: Virginia Tech and Virginia Tech and Duke University
• Keywords: Propensity score; Causal Inference; Driving behavior; Crash risk; case-cohort study

Abstract

This paper aims to demonstrate how to appropriately reduce the confounding effect in observational study step by step in a context of evaluating the causal effect of cellphone-using while driving on crash risk using case-cohort data from the Second Strategic Highway Research Plan Naturalistic Driving Study (SHRP2 NDS). Around 1,000 crashes and 20,000 control driving segments are used to represent the exposure status. In this paper, we use propensity score weighting to adjust the crash risk odds ratio between cellphone engaged situation and undistracted driving situation. First, we discuss the estimation of propensity score in the case-cohort study and its potential issues. Second, we evaluate the covariate balance between treated subject and untreated subjects after we adjust the data by propensity score. Third, we use three different propensity score weighting to estimate odds ratio of cellphone use on crash risk by age group.