I have three dichotomous nominal independent variables and one four-level nominal dependent variable. The dependent variable is measured multiple times (either 3 or 4, randomly determined) per configuration of IV conditions. All data is within-subjects; that is, every subject experiences each configuration of IV conditions exactly once.

  1. Is it even possible to run a multinomial regression given the repeated measures and multiple measurements? Or should I instead dummy code the DV into three binary variables, and either run a series of binary logistic regressions or average each of the binary variables across each config of conditions and run a multivariate multiple linear regression? (For the record, I have actually performed this dummy coding, and it results in a significant amount of missing data due to the rarity of two of the four DV levels - at least half the subjects would need to be excluded. I'd like to avoid this if possible - are there any alternatives?)

  2. If I should run logistic regressions (either a single multinomial regression or a series of binary regressions), are there any resources for how to run such a test in common statistical software packages (R, SPSS, SAS, HLM, etc.) that touch on analyzing repeated measures data with multiple IVs and multiple measurements per condition?


1 Answer 1


If your DV has 4 levels, then use multinomial logistic regression rather than set of binary logistic ones. In R, the excellent "lme4" package should allow you to handle the panel nature of your data (I think the command you need to use is "glmer").


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.