Model check after GEE

[Miyuki Takase]

Hello,

I run GEE using the following codes,
. xtset id
. xtgee judge i.sex age i.edu i.specialty i.transfer i.cog_chara i.urination i.route i.med i.agep i.bn_d i.bb_d i.ab_d, corr(exchangeable) nmp vce(robust)

As you can see, most of the predictor variables are categorical or binary, except for “age”. I also used vce(robust), because the multivariate normality test on the dependent variable indicated non-normality. Then, I checked the residuals using the following codes,

predict pred, xb
predict sd, stdp
gen res=(judge-pred)
summ res
gen stres=res/r(sd)
scatter stres pred, yline(0) mlab(id) mlabpos(0)
hist stres
sktest stres

I obtained the following results, which suggest the violation of assumptions.
Graph1.gph
Graph2.gph

Skewness and kurtosis tests for normality
----- Joint test -----
Variable | Obs Pr(skewness) Pr(kurtosis) Adj chi2(2) Prob>chi2
-------------+-----------------------------------------------------------------
stres | 5,814 0.0000 0.1286 71.56 0.0000

Could anyone please suggest possible solutions?

Thanks in advance.
Regards,
Miyuki