Robust polytomous logistic regression

In the context of polytomous regression, as with any generalized linear model, robustness issues are well documented. Existing robust estimators designed to protect against misclassification, but do not outlying covariates. It is shown that this can have a much bigger impact on estimation and testing than misclassification alone. To address problem, two new introduced: model-type estimator an optimal B-robust estimator, together corresponding Wald-type score-type tests. Asymptotic distributions variances these provided asymptotic test statistics under null hypothesis. A complete comparison proposed existing alternatives presented. This performed theoretically by studying influence functions estimators, empirically through simulations applications medical dataset.