Dealing with multicollinearity and outliers in multinomial logit model: a simulation study.

Multinomial logit model is a powerful tool for modeling the dependence relationship between a set of quantitative regressors and a categorical response variable with more than two categories. However, it necessitates careful consideration of multicollinearity among the regressors and examination for outliers. For dealing with these problems in the multinomial logit model, an approach based on robust principal components has been developed. The robust approach has achieved satisfactory results on real data. This paper aims to validate such an approach by a simulation study.