Using graphical models to interpret pavement condition data

Most of the multiple linear regressions developed in pavement performance studies have one response variable and one block of several explanatory variables. In this paper, graphical chain models are used to develop blocks of several possibly interacting responses with each block containing several interacting variables. The graphical chain model provides easy interpretation of conditional independence structure within the blocks. This ultimately provides an idea about association and dependence of different pavement condition variables. An example was presented using data from the Highway maintenance and Design Model (HDM).