My data is ordinal, and so missing values are imputed with the polr method from the MICE package. Now I have multiple datasets which I can run an Ordinal Logistic Regression on. But, as the title mentioned: I want to perform a Brant test to check the parallel regression assumption. How can I perform such a test on my imputed datasets?

olr <- with(imputed, polr(target ~ var1+var2)) 
olrsummary <- summary(pool(olr))

> brant(olr)
Error in formula.default(model) : invalid formula
> brant(olrsummary)
Error in temp.data[, name] : incorrect number of dimensions

I know I can take the first dataset with complete(imputed, 1) and use that for my Brant test. But that just don't sees right.


Without a better description of the data you use, it is difficult to get an idea of what is causing your error. However, you can try the following:

  1. Verify that the target variable is an ordinal variable.
  2. Consider the possibility that you are using a buggy version of the brant package. I faced a similar problem running the function with my own dataset and surprisingly the problem disappeared when I forked the project and ran it as my own script.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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