Standard and Shuffled Halton sequences in a Mixed Logit Model

摘要

Modeling consumer choice in different areas has lead to an increase useof discrete choice models. Probit or Multinomial Logit Models are oftenthe base of further empirical research of consumer choice. In some of thesemodels the equations to solve have no closed-form expression. They includemulti-dimensional integrals which can not be solved analytically.Simulation methods have been developed to approximate a solution forthese integrals. This paper describes the Standard Halton sequence anda modification of it, the Shuffled Halton sequence. Both are simulationmethods which can reduce computational effort compared to a randomsequence. We compare the simulation methods in their coverage of themulti-dimensional area and in their estimation results using data of consumerchoice on grocery store formats.