Difference between default SE and robust SE in generalized estimating equations?

#1
As topic; Just started to learn more about GEE and I know that two different possible approaches dealing with standard error (SE) when fitting generalized estimating equations models would be standard (default) SE and "robust" SE. What are the differences between these two approaches and also, are there any advantages and disadvantages of each method?
 

hlsmith

Omega Contributor
#2
Sorry for the lack of detail or citations, robust are typically larger and may be related to construction of sandwich estimators (Huber-White), or depending on the context you are looking at may reflect covariance (since some models control for between and within group effects).
 

noetsi

Fortran must die
#3
White robust SE typically deal with problems in SE caused by heteroscedastcity. That is they correct for the problems caused by that. Some recommend them for all analysis (I no longer remember why).