Statistical models have been very popular for estimating the performance of highway safety improvement programs which are intended to reduce motor vehicle crashes. The traditional Poisson and Poisson-gamma (negative binomial) models are the most popular probabilistic models used by transportation safety analysts for analyzing traffic crash data. The Poisson-gamma model is usually preferred over traditional Poisson model since crash data usually exhibit over-dispersion. Although the Poisson-gamma model is popular in traffic safety analysis, this model has limitations particularly when crash data are characterized by small sample size and low sample mean values. Also, researchers have found that the Poisson-gamma model has difficulties in handling under-dispersed crash data. The primary objective of this research is to evaluate the performance of the Conway-Maxwell-Poisson (COM-Poisson) model for various situations and to examine its application for analyzing traffic crash datasets exhibiting over- and under-dispersion. This study makes use of various simulated and observed crash datasets for accomplishing the objectives of this research. Using a simulation study, it was found that the COM-Poisson model can handle under-, equi- and over-dispersed datasets with different mean values, although the credible intervals are found to be wider for low sample mean values. The computational burden of its implementation is also not prohibitive. Using intersection crash data collected in Toronto and segment crash data collected in Texas, the results show that COM-Poisson models perform as well as Poisson-gamma models in terms of goodness-of-fit statistics and predictive performance. With the use of crash data collected at railway-highway crossings in South Korea, several COM-Poisson models were estimated and it was found that the COM-Poisson model can handle crash data when the modeling output shows signs of under-dispersion. The results also show that the COM-Poisson model provides better statistical performance than the gamma probability and traditional Poisson models. Furthermore, it was found that the COM-Poisson model has limitations similar to that of the Poisson-gamma model when handling data with low sample mean and small sample size. Despite its limitations for low sample mean values for over-dispersed datasets, the COM-Poisson is still a flexible method for analyzing crash data.

In recent actuarial literature, the bivariate Poisson regression model has been found to be useful for modeling paired count data. However, the basic assumption of marginal equi-dispersion may be quite restrictive in practice. To overcome this limitation, we consider here the recently developed bivariate Conway–Maxwell–Poisson (CMP) distribution. As a distribution that allows data dispersion, the bivariate CMP distribution is a flexible distribution which includes the bivariate Poisson, bivariate Bernoulli and bivariate Geometric distributions all as special cases. We discuss inferential methods for this CMP distribution. An application to automobile insurance data demonstrates its usefulness as an alternative framework to the commonly used bivariate Poisson model. / Thesis / Master of Science (MSc)

Indiana University-Purdue University Indianapolis (IUPUI) / The two-part model (2PM) is the most widely applied modeling and estimation framework in empirical health economics. By design, the two-part model allows the process governing observation at zero to systematically differ from that which determines non-zero observations. The former is commonly referred to as the extensive margin (EM) and the latter is called the intensive margin (IM). The analytic focus of my dissertation is on the development of a general framework for specifying, estimating and drawing inference regarding causally interpretable (CI) effect parameters in the 2PM context. Our proposed fully parametric 2PM (FP2PM) framework comprises very flexible versions of the EM and IM for both continuous and count-valued outcome models and encompasses all implementations of the 2PM found in the literature. Because our modeling approach is potential outcomes (PO) based, it provides a context for clear definition of targeted counterfactual CI parameters of interest. This PO basis also provides a context for identifying the conditions under which such parameters can be consistently estimated using the observable data (via the appropriately specified data generating process). These conditions also ensure that the estimation results are CI. There is substantial literature on statistical testing for model selection in the 2PM context, yet there has been virtually no attention paid to testing the “one-part” null hypothesis. Within our general modeling and estimation framework, we devise a relatively simple test of that null for both continuous and count-valued outcomes. We illustrate our proposed model, method and testing protocol in the context of estimating price effects on the demand for alcohol.

In traffic safety analysis, a large number of distributions have been proposed to analyze motor vehicle crashes. Among those distributions, the traditional Poisson and Negative Binomial (NB) distributions have been the most commonly used. Although the Poisson and NB models possess desirable statistical properties, their application on modeling motor vehicle crashes are associated with limitations. In practice, traffic crash data are often over-dispersed. On rare occasions, they have shown to be under-dispersed. The over-dispersed and under-dispersed data can lead to the inconsistent standard errors of parameter estimates using the traditional Poisson distribution. Although the NB has been found to be able to model over-dispersed data, it cannot handle under-dispersed data. Among those distributions proposed to handle over-dispersed and under-dispersed datasets, the Conway-Maxwell-Poisson (COM-Poisson) and double Poisson (DP) distributions are particularly noteworthy. The DP distribution and its generalized linear model (GLM) framework has seldom been investigated and applied since its first introduction 25 years ago. The objectives of this study are to: 1) examine the applicability of the DP distribution and its regression model for analyzing crash data characterized by over- and under-dispersion, and 2) compare the performances of the DP distribution and DP GLM with those of the COM-Poisson distribution and COM-Poisson GLM in terms of goodness-of-fit (GOF) and theoretical soundness. All the DP GLMs in this study were developed based on the approximate probability mass function (PMF) of the DP distribution. Based on the simulated data, it was found that the COM-Poisson distribution performed better than the DP distribution for all nine mean-dispersion scenarios and that the DP distribution worked better for high mean scenarios independent of the type of dispersion. Using two over-dispersed empirical datasets, the results demonstrated that the DP GLM fitted the over-dispersed data almost the same as the NB model and COM-Poisson GLM. With the use of the under-dispersed empirical crash data, it was found that the overall performance of the DP GLM was much better than that of the COM-Poisson GLM in handling the under-dispersed crash data. Furthermore, it was found that the mathematics to manipulate the DP GLM was much easier than for the COM-Poisson GLM and that the DP GLM always gave smaller standard errors for the estimated coefficients.