I'm running a multiple logistic regression model which includes as independent variables continuous, nominal dichotomous and polytomous variables (to be precise: 2 continuous, 4 nominal dichotomous and 1 polytomous). As dependent variable i use the clinical outcome at the end of follow-up. The polytomous variable refers to 6 different diagnostic subgroups; I wouldn't include this variable in the model using dummy coding because there isn't any reference group and i would like to explore the role of each diagnostic group on the clinical outcome. Can anyone help me to solve this problem?


1 Answer 1


It seems you would like to represent a polytomous (6 levels) variable with all its levels, since there is no (natural) reference level. But that would lead to an overparameterized model (unless you leave out the intercept, which is maybe not natural when you have many factor variables).

Your problem is that you want to interpret the results only by looking at a standard summary output table. But that is a very limited way to interpret model results. When you have an estimated model, use it to

  • test parameter contrasts that are of interest to you

  • Visualize the model!

Some similar posts:
Interpreting coefficient in GLM with categorical explanatory variables,

Why is it necessary to "ignore" a level when applying sum contrasts?,

Possible to code contrasts comparing each level to grand mean with no reference category?,

What to do in a multinomial logistic regression when all levels of DV are of interest?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.