Sampling and discrete choice

Sampling is a key step in the empirical study of choice behaviour. It consists of selecting observations and/or alternatives from a target population within an available resource budget under a specified protocol. In this chapter, we review several sampling approaches and discuss their implications for choice model parameter estimation and the use of an estimated model for prediction. Specifically, we focus on the sampling of observations (which is needed when the size of the population is large) using exogenous and endogenous variables and the sampling of alternatives (which is needed when the number of alternatives in the studied choice situations is large). We also consider the prediction of market shares and the calculation of elasticities under different sampling approaches. The sampling concepts are illustrated using semi- and fully-synthetic data. Finally, the chapter concludes with a brief discussion of additional literature of sampling approaches and applications of sampling in choice analysis.