Hi!
I have a linear mixed model for repeated measures where the residuals are correlated and there is some heteroscedasticity present (due to time). It seems to me that when choosing residual structure, you have to choose whether to account for the correlation or for the heteroscedasticity, but that you cannot account for both at the same time (ideally, I think I would have to have something like res(exp, t(time) by(time)), but that is not allowed). I have heard about auto-regressive heterogeneous models in R. Is there something similar in Stata?
Kjell Weyde
I have a linear mixed model for repeated measures where the residuals are correlated and there is some heteroscedasticity present (due to time). It seems to me that when choosing residual structure, you have to choose whether to account for the correlation or for the heteroscedasticity, but that you cannot account for both at the same time (ideally, I think I would have to have something like res(exp, t(time) by(time)), but that is not allowed). I have heard about auto-regressive heterogeneous models in R. Is there something similar in Stata?
Kjell Weyde
