2016/02/18 by Kersten, Norbert
Medicine · #610 Medical sciences #Medicine #categorial outcome variables #kategoriale Zielvariablen #logistic regression model #logistisches Regressionsmodell #multinomial #multinomiale #ordinal and stereotype regression #ordinale und stereotype Regression
paper · doi:10.3205/mibe000163
For about three decades now a number of suitable regression models for categorical outcome variables have been described in the literature. These are also available as analysis tools in statistical software packages. This overview presents some of the models which are suitable for medicine and health sciences. The lesser-known stereotypical model in particular is emphasized since the ordinal and the multinomial models have been in use for some time. The article is aimed at statistics users: It covers types of categorical data, models and their implementation in statistical software. Estimation and inference theory is omitted. It demonstrates, firstly, model selection depending on the order structure of the outcome variables and, secondly, appropriate interpretation of the model parameters. This is illustrated by one example each for the multinomial and the stereotypical model. Notes to use the software are included.