Analysis of Multinomial Discrete Choice Probit Model with Structured Correlation Matrix

• Activity Number: 617
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Bhaskara Ravi*+ and N. Rao Chaganty
• Companies: Old Dominion University and Old Dominion University
• Keywords: Discrete Choice ; Multinomial Probit ; Equicorrelation ; Stochastic representation

Abstract

In Econometrics, McFadden's(1974) Conditional Logit(CNL) model is the most widely used Discrete Choice Model for analyzing choice behavior. This model is based on the assumptions that the unobserved factors are independent and follow a Gumbel distribution, but these assumptions may not be appropriate in some situations. An alternative model that allows dependency between the unobserved factors is the "Multivariate Discrete Choice Probit (MDCP) Model". Note that this model involves numerical evaluation of multivariate normal distribution function to compute the choice probabilities. In this talk, I will discuss exact analytical expressions for the choice probabilities assuming an equicorrelated structure using a stochastic representation. I will also present the procedure of obtaining the maximum likelihood estimates with analytical expressions for the Fisher information matrix to compute standard errors. Further, I will show that the MDCP model is more efficient than the CNL model asymptotically and in small samples with use of SAS and R programs.