1

I have a question regarding the homogeneity of variance in three regression models of diffrent datasets belonging to the same multiple imputed data.

As I used multiple imputation I have to check regression assumtions on single datasets. The majority of the Datasets, belonging to the same model, pointing in the direction that homogeneity of variance is not given. The NCV and Breusch-Pagan test both indicated p values <0.05, thus rejected the null hypothesis of constand variances. However, from looking at the std. residual vs. fitted vaule plots, (https://i.stack.imgur.com/4uB3O.jpg ) I would assume the variances as roughly okay (cf. the attached picture), but I am certainly not an expert.

The problem is also that if the homoscedasticity assumption indeed would be violated and I wanted to correct for heteroscedasticity (e.g. by using the sandwich paket and calculating robust std erros) , I would have to do that in a pooled manner - on all the multiple imputed data sets, and I could not find a R funtion which adresses that kind of problem. I already tried out Box-Cox transformation for the dependet variable which did not resolve the problem. Maybe one of u has an idea how to adress this issue?

Thanks for ur help

Best wishes

Anna

Ecidem
  • 31
  • 1
  • Cross validated: https://stats.stackexchange.com/questions might be a better place for your question as it is not strictly coding-related. – Shibaprasadb Sep 08 '21 at 08:40
  • Ah, you're right. Thank you for that hint. I was already confused that I could not find related tags here. – Ecidem Sep 08 '21 at 09:43

0 Answers0