I'm running HLM in SAS using proc mixed - students nested within schools. I'm thinking that there is heteroskedasticity in two of the schoollevel variables but I'm not sure how to confirm. That is, when two of the independent variables are at low levels, the outcome variable has little variance, but there is more variability when those predictor variables increase. I'm not sure how to correct for it. Is it possible to randomize the school level variables??? I cannot enter school level variables into the random statement in SAS.
I could make an indicator/dummy variable for the low values on those two predictor variables and then leave the continuous variables in the model as well. Or, I could categorize each of the variables, ensuring that the "low" category is captured. Any advice is much appreciated.
Thanks for your help,


雷达卡


京公网安备 11010802022788号







