Graduate Thesis Or Dissertation

An Evaluation of the Random-Parameter and Latent Class Methods for Heavy Vehicle Injury Severity and Crash Rate Analysis : An Idaho Case Study by Roadway Classification

Public Deposited

Downloadable Content

Download PDF


Attribute NameValues
  • This thesis provides a comparison of advanced econometric frameworks to account for unobserved factors in crash reported data (also referred to as unobserved heterogeneity) while identifying contributing factors by roadway classification for heavy vehicle injury severity and crash rates. The presented thesis provides two manuscripts that expand the literature regarding these advanced econometric methods using Idaho heavy vehicle crash reported data as a case study.The first manuscript utilizes two advanced analytical techniques, namely the random-parameter multinomial logit (also referred to as the mixed logit) and latent class logit, to identify injury severity contributing factors while exploring the empirical results of the two methods. Recent efforts suggest that more studies examining the results of the two approaches be completed to facilitate the identification of a superior framework that can used for future analyses. In comparing overall model fit (log-likelihood values), marginal effects and actual severities versus predicted severities, it was found that the latent class framework for heavy vehicle injury severity analysis performed better for the Idaho crash data. Further, through a model separation test, it was found that road classifications need to be analyzed separately with 99.99% confidence.In regard to the second manuscript, two additional advanced econometric approaches were utilized to investigate the factors that contribute to the number of crashes per million-vehicle-miles-traveled. Again, analysis was completed by road classification, as it was discovered in manuscript one that road classifications need to be analyzed separately. Due to the skewed distribution of heavy vehicle crash rates, Tobit regression was applied and compared to the empirical results of a latent class Tobit regression framework. To determine the most statistically significant method, overall model fit, partial effects and actual crash rates versus predicted crash rates were evaluated. The latent class Tobit regression framework outperformed that of the traditional Tobit regression approach for the Idaho dataset.Through the comparison of the crash analysis framework, latent class logit and latent class Tobit regression were found to outperform their traditional counterparts. In the midst of evaluating the empirical results, this thesis has statistically determined that road classifications need to be analyzed individually. The current thesis extends the literature in regard to heavy vehicle injury severity analysis and fills the noticeable gap that exists for heavy vehicle crash rate analysis. An analytical foundation has been provided and can be used for future studies that need to model discrete outcomes or continuous response variables. Although agencies typically do not use such advanced methods, the results from this thesis can help the Idaho Department of Transportation facilitate crash countermeasures with more precision and allow them to prioritize accordingly.
  • Keywords: Tobits, Mixed Logit, Tobit, Unobserved Heterogeneity, Roadway Classification, Crash Rate, Heavy Vehicle, Injury Severity, Random-Parameter, Latent Class
Resource Type
Date Available
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Committee Member
Academic Affiliation
Non-Academic Affiliation
Rights Statement
Peer Reviewed



This work has no parents.

In Collection: