Confounder confounds at only one level of polytomous variable


New Member
Hello! I have looked for a post on this but have been unsuccessful. It would be great if you could help.

I am using logistic regression in SAS.

I have a polytomous variable, race, with 5 levels. When I add covariate A, the coefficient of only 1 level of race changes significantly (by 44%). This level was not significant to begin with.

Should I add covariate A to the model, since (a) it only affects 1 level of race and (b) that 1 level was not significant in the model?

Thank you for your help!