Hi Rashed, I'm think that you are trying to say that you could use gee to create mixed effect model with a random effects term of race to handle intra-race/ethnicity correlations. That is an interesting idea, but I have never seen this used. Do you have references to biostatistical papers alluding to this? Does this model allow for creating a precise measurement for interaction on the additive or multiplicative scale (ie, does the stata output generate a terms for this interaction?)
Thanks for sharing!
Asa