Impact of A Priori Distributions on Mixed Logit Model Estimation Tests on Synthetic Data
Majken Vildrik Sørensen, Otto Anker Nielsen · Technical University of Denmark, DTU Orbit (Technical University of Denmark, DTU) · 2004
This paper focuses on the MNL and the Mixed Logit models’ ability to correctly estimate a model. Synthetic data has been formed followed by estimation of MNL and Mixed Logit models conditional of a variety of shapes of distribution for distributed terms. Model selection based solely on a combination of MSL estimation, t-test and likelihood ratio test is shown to be ambiguous. A test for determining the shape of distribution is proposed. Application of the method retrieved the distribution from data.