Multinomial logistic regression is an extension of this approach to situations where the response variable is categorical and has more than two possible values.
©FSRH J Fam Plann Reprod Health Care 2008: 34(3) What is it? When a response variable has only two possible values (e.g. recurrence/not), binary logistic regression is commonly used to test or model the association between that response and a number of potential explanatory variables, with each association estimated in terms of an odds ratio (OR). Multinomial logistic regression is an extension of this approach to situations where the response variable is categorical and has more than two possible values. Ordinal logistic regression is a special type of multinomial regression, which can be advantageous when the response variable is ordinal. [See Box 1 for glossary of terms.]