I have a dataset where students are nested within classes, and classes are nested within schools. I am interested in evaluating the effects of a treatment delivered at the student level. The response variable is on a continuous scale. The residuals with a Linear Mixed Model (with classes and school as random factors) are not normally distributed (they show left skewness). Similarly for the random effects.

If this was a 2 level setting, the lack of normality would not be an issue with a Generalized Estimation Equation (GEE). However, I have classes also nested within schools (i.e., a third level of nesting).

Do you know any reference indicating whether GEEs can also be applied in 3 level settings?

What else may I do asides from data transformations or multi-level Boostrap if GEEs are not applicable?


1 Answer 1


In case of a continuous outcome, the GEE is almost equivalent to fitting a mixed model and using the sandwich estimator for the standard errors; for more on this, check this answer.

You have not given us any information on what exactly the problem is with the residuals. You could, for example, consider fitting a mixed model with another distribution for your outcome variable.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.