(Reposting since I didn't get any responses last week.)
When using categorical covariates in HB analysis for MaxDiff/CBC, is there a minimum sample size that should exist in each variable value?
For example, if the total sample size is 500 respondents, and I have a categorical variable indicating 10 unique groups of 50 respondents each, would it be wise/unwise to use the covariate? I know 50 respondents is well below the very general rule of thumb of 200 respondents per audience, so it might not be a good idea to look at the results summarized by audience, but would the covariate still be useful in assisting the analysis, or could it be detrimental?