Evaluation of a new mixed logit (MIXL) estimator using Monte Carlo simultations

Summary

The random parameter logit (RPL) or mixed logit (MIXL) model is the most commonly used model to estimate respondents' preferences obtained using discrete choice experiments (DCEs). However, currently available estimators either produce unreliable results or are prohibitively slow when used with larger datasets, which warrants investigations into a more reliable and faster approach to analyzing large-scale DCEs. A promising candidate would be an estimator based on Laplace Approximations. In this application, I request the required computation time to establish the relative performance of Laplace Approximations versus existing MIXL estimators using the Cartesius computing facilities.

Details

Project number

16682

Main applicant

Dr. M.F. Jonker

Affiliated with

Erasmus Universiteit Rotterdam, Erasmus MC, Maatschappelijke Gezondheidszorg

Duration

01/12/2017 to 01/01/2018