The Statistical Relationship between Bivariate and Multinomial Choice Models
The authors demonstrate the conditions under which the bivariate probit model can be considered a special case of the more general multinomial probit model. Since the attendant parameter restrictions produce a singular covariance matrix, the subsequent problems of testing on the boundary of the parameter space are circumvented by the construction of a score test.
discrete choice, bivariate probit, multinomial probit, nested choice models
Faculty of Economics