Multinomial Logistic Regression

This chapter discusses the logit link function of logistic regression to the multinomial situation where a categorical response variable can take on one of several outcomes. It talks about an approach of estimating the class probabilities for a multicategory response, and uses these probabilities to classify new cases into one of several outcome groups. Several choices are available to estimate multinomial logistic regression models in R. For example, one can use the command mlogit in the package mlogit, the command vglm in the package VGAM, or the mnlm function in the package textir. The chapter illustrates an example: forensic glass. The three features, proportions of Na, Mg, and Al, are used to illustrate the multinomial logistic regression model. The R function mnlm makes use of simple triplet matrices.