On the performance of the semiparametric binary response model when the true model is parametric logistic

Abstract

In this article, a simulation study is performed to reveal the devia- tions of the semiparametric binary response model from its parametric counterpart, based on various scenarios including different sample sizes, different bandwidth parameters containing the optimal ones, different forms of the linear index function and two and higher dimensional cases of the explanatory variables when the true model is logistic regression. The method of the Density Weighted Average Derivative Estimator (DWADE) is used in the semi-parametric estimation. A real data set on liquefaction is used to demonstrate the effectiveness of the simula- tion results with the results in practice. Additionally, new commands written for the estimation of both models in the Windows based 4.8 version of the XploRe package are introduced. This study may be seen as an updated form of the article by Proen¸ca andWerwatz (1994) which used XploRe commands written for both estimators in an old MS Dos format.