ABSTRACT
Traditional modeling approaches treat estimated parameters as constant across observations. That is, the effect of any individual explanatory variable (in the X vector) is the same for each observation or individual. However, this fixed-parameter assumption may be incorrect because of unobserved factors that may vary from one observation/individual to another. To account for these unobserved factors, or unobserved heterogeneity, random parameters models are derived by assuming that the estimated parameters vary across observations, usually according to some pre-specified distribution. Accounting for the possibility that parameters may vary significantly across observations makes model estimation considerably more complex because a mixing distribution is introduced into the estimation process. The motivation for such models is to account for unobserved heterogeneity across observations and, by doing so, potentially make important new insights regarding the analysis of the data. This chapter presents the logic behind random parameters models and their application to different types of data, including discrete, count, and continuous data. Grouped and correlated random parameters formulations are also discussed and numerous examples are provided. This chapter demonstrates the great potential of such model in transportation data analysis.
