Advanced discrete choice models—in particular, mixed logit models—are used extensively in transportation. Although much progress in estimation techniques has made them numerically appealing, their properties have not been fully explored. This lack of exploration sometimes leads to confusing quality measurements and misinterpretation of the estimates. In this paper, the regularity conditions for which the information equality holds are reviewed, and some underlying technical difficulties in the context of mixed logit modeling are discussed. This paper specifically addresses the questions of correlations between estimated parameters and the validity of the asymptotic normality assumption in complex models, as nonparametric formulations. In the latter case, the population is resampled with the use of bootstrap principles to construct confidence intervals on the estimated parameters. Numerical tests on simulated data are presented to assess the relevance of the problem and the validity of the methods proposed.
On the Information Matrix in Mixed Logit Models Estimation
Transportation Research Record: Journal of the Transportation Research Board
Transportation Research Record: Journal of the Transportation Research Board ; 2254 , 1 ; 11-18
2011-01-01
Article (Journal)
Electronic Resource
English
On the Information Matrix in Mixed Logit Models Estimation
Online Contents | 2011
|British Library Conference Proceedings | 2005
|Estimation of Value of Time Using Mixed Logit Models
British Library Conference Proceedings | 2005
|Estimation of Logit Choice Models Using Mixed Stated-Preference and Revealed-Preference Information
British Library Conference Proceedings | 1997
|Optimal design for mixed logit models
British Library Conference Proceedings | 2005
|