共役性概念に基づくロジットモデルのパラメータ推定法とその統計的検定について

抄録

This paper aims at providing disaggregate tests theoreticaly consistent with a parameter estimation method. The parameter estimation method presented here is based on a straightforward application of the conjugacy theory on the statisfaction function of disaggregate logit model. Because the conjugacy theory indicates that the optimum function of conjugate problem induces an entropy of choice behavior, we can use the entropy as a measure of uncertainty involved in estimating of parameters. Then, it can be show that Shannon's information can be derived from the entropy as the conjugate function of satisfaction function and that Shannon's information is useful for comprehensive tests such as “significance”, “usefulness” and “accuracy” of models.