Robust estimation of qualitative response regression models

Abstract

Qualitative response regression models such as logistic regression are typically estimated by the maximum likelihood method. To improve its robustness, two special cases of the M-estimation based approach for quantitative continuous random variables were extended to the variant of qualitative and mixed variables modeling. Expressions of the score functions for polytomous regression models were derived. In according to results of the research some conclusions and practical recommendations were given.

Publication
In Proceedings of the International Workshop “Applied Methods of Statistical Analysis. Simulations and Statistical Inference”, Novosibirsk, Russia