If an outcome variable takes only a few discrete values, it may be appropriate to treat it as categorical data.
If there are only two categories present, the outcome will be estimated with a logistic model (also known as logit), probit model, or linear model. If more than two categories are present, the outcome can be estimated with either a multinomial logit model (appropriate for unordered categories, for example, red, green, and blue) or an ordered probit model (appropriate for ordered categories, for example, never, sometimes, and always).
To treat an outcome variable as categorical data:
The desired model can then be chosen using the controls below the outcome table. When categorical outcomes are estimated with a multinomial logit model, a separate set of coefficients is estimated for every value of the outcome. In contrast, with the other models, a single set of coefficients applies to all category values.
Note that treating an outcome as categorical data in the Model view does not affect whether the underlying column is treated as categorical data in the Summary view.