Estimating Driver Crash Risks Based on the Extended Bradley–Terry Model: An Induced Exposure Method

Abstract

Quantifying driver crash risks has been dif®cult because the exposure data are often incompatible with crash frequency data. Induced exposure methods provide a promising idea that a relative measurement of driver crash risks can be derived solely from crash frequency data. This paper describes an application of the extended Bradley±Terry model for paired preferences to estimating driver crash risks. We estimate the crash risk for driver groups de®ned by driver±vehicle characteristics from log-linear models in terms of a set of relative risk scores by using only crash frequency data. Illustrative examples using police-reported crash data from Hawaii are presented.

Driver injury severity: an application of ordered probit models

“…Zhang et al (2000) find the elderly to be at increased risk for fatal injury, and Li and Kim (2000) find the elderly to be more often at fault. In terms of health effects on crash rates, Hu et al (1998) have looked critically at medical conditions of drivers and their physical disabilities.…”

Using Bradley–Terry models to analyse test match cricket

“…Popular applications include analysing taste test experiments (Hopkins (1954), Bliss et al (1956)) and modelling sports competitions, with racquetball (Strauss & Arnold (1987)), soccer (Hallinan (2005)) and basketball (Cattelan et al (2013)) all receiving attention in this regard. More creative examples of applications include statistical modelling of citation exchange among statistics journals (Varin et al (2013)), predicting the fighting ability of lizards (Whiting et al (2006)) and estimating driver crash risks (Li & Kim (2000)).…”

The efficiency of using non-culpable crash-claim involvements from insurance data as a means of estimating travel exposure for road user sub-groups

“…In spite of deeming that their data appeared to confirm a major quasi-induced exposure assumption (namely that the non-responsible proportions were similar for different state-oflicensure groups) they concluded that quasi-induced exposure had limitations that made it unsuitable for estimating the numbers of unlicensed drivers on California roads. Cuthbert's (1994) formulation for extending the induced exposure concept was utilized by Redondo-Calderon et al (2001) and expanded by Li and Kim (2000) using a log-linear model for paired preferences in assessing crash risk for driver age/gender combinations. Both studies used data that it was claimed distinguished reliably between high-and low-risk driver/crash characteristic groupings, and both produced results that supported the applicability of the induced exposure concept where comparisons to travel survey data or expectations with respect to conventional exposure information were considered.…”