Sample procedures for the statistical and empirical analyses of probabilistic discrete decision models

This paper establishes the procedures for performing statistical analyses of and empirical comparisons among probabilistic discrete decision models with simulated data. The generation of information by means of a novel technique in this area, namely assignment of responses by multinomial simulation, is explained. Estimation to parameters, tests of hypotheses, and goodness of fit measures, with special reference to maximum likelihood methods, are examined. The applicability of the procedures to the study of the Multinomial Logit Model and two parametric generalizations of this model, the Generalized Extreme Value Model and the Nesting Extreme Value Model in the three alternative, two substitute situation is studied. An empirical investigation carried out is outlined.