Survey data, collected to calibrate discrete choice models, are often with partially missing information on some attributes. Ways to fill in missing data are needed in order to apply standard calibration procedures; however, this may result in seriously biased estimates. In this paper, we propose a calibration approach that takes advantage of the information of the observed attributes as well as to infer the missing data statistics. In doing so, we show in a simulated dataset that the calibration can produce fairly good estimate and correctly account for tendency of bias between users and non-users.
Discrete Choice Modeling with Partially Missing Information from a Revealed Preference Survey
Ninth International Conference on Applications of Advanced Technology in Transportation (AATT) ; 2006 ; Chicago, Illinois, United States
2006-08-04
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Discrete Choice Modeling with Partially Missing Information from a Revealed Preference Survey
British Library Conference Proceedings | 2005
|Estimation of Logit Choice Models Using Mixed Stated-Preference and Revealed-Preference Information
British Library Conference Proceedings | 1997
|Revealed Preference Analysis of Commercial Vehicle Choice Process
Online Contents | 2002
|Revealed Preference Analysis of Commercial Vehicle Choice Process
British Library Online Contents | 2002
|Preference endogeneity in discrete choice models
Online Contents | 2014
|