Exploring the indirect inference estimator (137)
Project Number: 137
Project Leader: Marcus Sundberg
Start Date: 2010-08-31
End Date: 2010-12-31
Description:Estimation of route choice models in a discrete choice framework poses a number of problems. In the literature there are basically two different kinds of econometric model (i) logit models with different correction terms to deal with correlation across routes and (ii) explicit modeling of correlation structure by assigning random link costs, often in the context of probit modeling approaches. Indirect inference was put forward as a promise that almost any model can be estimated, as long as you can simulate it. In this context, indirect inference may be seen as a formalized method to do what is heuristically already done. But in doing so, one is able to control for the bias introduced by using a misspecified model. In this project we explore how indirect inference can be used in the context of, for instance, mode- and destination choice or route choice.