Hello,

I am running multilevel models in STATA, including multiple predictor variables as both fixed and random effects. I need to use bootstrapping for the standard errors in the models because (by nature of the data) the variables are highly skewed. I have tried many transformations and non-parametric statistics. However, the transformations do not normalize the data and my team is insisting on using MLM because of the nested nature of my data.

I am able to run bootstrapping for my fixed effects predictors using the command " bootstrap _b _se, reps(50) : xtmixed " in the do file. However, when I try to run bootstrapping after adding predictors as random effects, the model seems to get stuck (i.e., not run), or if it does run it only gives boostrapped estimates for the fixed effects and none for the random effects.

Any thoughts on how to run such a model?