0

I'm running Linear Mixed Models on a dataset. The assumption for homoscedasticity is not being met, however when I remove one independent variable, then it's being met. So all the other variables except this one are homoscedastic. To fix this and make the dataset fit the model better, could I just transform the independent variable which is the issue by cube rooting it? And leaving others as they are? Or do I have to transform all of them if I'm transforming one variable?

Also, is transforming them going to increase any kind of error rates and make inferences difficult? I would really appreciate help with this and I apologize if this is a silly question, as I'm a beginner.

Anusha
  • 1
  • No, you don't have to transform all independent variables if you transform one. Yes, you need to consider the transformation when you look at your results. How exactly you need to do that depends on the transformation you used. – Sointu Apr 12 '23 at 14:56
  • Here is a good reply on interpreting cube root transformed variable's regression coefficient. – Sointu Apr 14 '23 at 10:32

0 Answers0