The Use Of On-Site Samples In Marketing Research

Abstract

On-site sampling procedures are frequently used in demand estimation applications.  This results in a sample that is both truncated and endogenously stratified, limiting the inferences that can be drawn from the sample.  Shaw (1985) develops a likelihood function that corrects for these problems and uses a Monte Carlo simulation to verify that the likelihood function outperforms models that ignore the problems.  However, the simulation is less than ideal.  In this paper, I develop an additional method for correctly for the problems associated with on-site samples and conduct a more thorough simulation to examine the performance of the various estimation methods.