Browsing by Subject "Unobserved heterogeneity"

Now showing 1 - 2 of 2

A comprehensive mixed logit analysis of crash type conditional on a crash event
(2015-12) Chu, Alice Ai-Ichi; Bhat, Chandra R. (Chandrasekhar R.), 1964-; Zhang, Zhanmin
This thesis presents a comprehensive mixed logit model of crash types, where the crash type outcomes are defined by a combination of the nature of collision and the types of vehicles involved in the crash. While prior research in the highway safety field has largely studied and modeled crashes along specific dimensions and categories, this study attempts to model the influence of various explanatory factors on crash type probabilities in a comprehensive and holistic way. The model considers 20 different crash types (alternatives) simultaneously. Using the 2011-2013 General Estimates System (GES) crash database in the United States, this research effort presents a mixed logit model that characterizes the effects of weather and seasonal variables, temporal attributes, roadway characteristics, and driver factors on the probability of observing various crash types. The model reveals the competing influences of various factors on different crash outcomes and the presence of significant unobserved heterogeneity in the manner in which variables affect crash type probabilities. The model offers a framework for developing safety measures and devices that do not result in unintended consequences where a reduction in one crash type probability is met with an increase in another crash type probability.
Modeling unobserved heterogeneity of spatially correlated count data using finite-mixture random parameters
(2015-05) Buddhavarapu, Prasad Naga Venkata Siva Rama; Scott, James (Statistician); Prozzi, Jorge A
The main goal of this research is to propose a specification to model the unobserved heterogeneity in count outcomes. A negative binomial likelihood is utilized for modeling count data. Unobserved heterogeneity is modeled using random model parameters with finite multi-variate normal mixture prior structure. The model simultaneously accounts for potential spatial correlation of crash counts from neighboring units. The model extracts the inherent groups of road segments with crash counts that are equally sensitive to the road attributes on an average; the heterogeneity within these groups is also allowed in the proposed framework. This research employs a computationally efficient Bayesian estimation framework to perform statistical inference of the proposed model. A Markov Chain Monte Carlo (MCMC) sampling strategy is proposed that leverages recent theoretical developments in data-augmentation algorithms, and elegantly sidesteps many of the computational difficulties usually associated with Bayesian inference of count models.