I am trying to find some resources on model validation for GEE-GLMs. Unfortunately I can't afford to purchase expensive textbooks and many of the books address GEEs using other software such as Stata.

Does anyone have good resources on the model validation process of GEE-GLMS particularly with respect to R packages that can be used? The Zuur textbook mentions that "The graphical validation uses Pearson residuals and follows the model validation steps of GLM." O know for GLMs we use residuals vs. fitted to check for homogeneity and a Q-Q plot of residuals to check for normality, but do we assess these same things in a GEE-GLM or are the assumptions different?

I haven't found R packages yet for how I would check the homogeneity and normality of residuals, and also how to plot the predictions of a GEE-GLM. Any help with this would be much appreciated!


1 Answer 1


The Generalized Estimation Equations (GEEs) are a semi-parametric approach that does not make any particular assumption for the distribution of your outcome variable. Also, the correlation structure you assume is a working correlation that is also not assumed to be correctly specified. The sandwich estimator takes care of this. Moreover, GEEs automatically also allow for over-dispersion.

You only need to appropriately specify the mean of your data by suitably specifying the link function and allowing for some flexibility in the functional form of the covariates, e.g., consider potential nonlinear terms.

Note, however, that the simple GEEs will only provide valid inferences under the missing completely at random assumption for any missing data you may have.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.