ABSTRACT
This chapter presents tools for illuminating relationships between variables in the data (including direct or causal relationships, indirect or mediated relationships, and associations) in the presence of measurement difficulties, endogeneity, and unobservable or latent variables. Examples of latent variables in transportation include attitudes toward transportation policies or programs such as gas taxes, van pooling, high occupancy vehicle lanes, and mass transit. Interest might also be centered on the role of education, socio-economic status, or attitudes on driver decision making, which are also difficult to measure directly. The chapter presents several approaches to uncovering data structure including principal components analysis (widely used as an exploratory method for revealing structure in data), factor analysis (a statistical approach for examining the underlying structure in multivariate data), and structural equation modeling (a framework developed specifically for dealing with unobservable or latent variables, endogeneity among variables, and complex underlying data structures encountered in social phenomena often entwined in transportation applications).
