st: xtmelogit and confidence interval for the random effect

["Song, Lin" <Lin.Song@kingcounty.gov>]

Dear Statalisters,

I have a question on how to calculate the confidence interval for the random effect after xtmelogit.

Using the BRSS data at the individual level for King County, my model is:

-xtmelogit obese var1 var2 || area:

Var1 is individual level data such age and race
Var2 is area level data such as poverty
Area: is the 48 health reporting areas in King County
After the model, I created the following post estimation variables and my purpose is to obtain the rate of obese and 95% CI at the area level:

-predict mu1, mu                             
-predict area_xb, xb
-predict area_re_m1, reffects    level(area) 
-predict area_re_s1, reses level(area)

-gen mu2 = 1 /(1+exp(-1*(area_xb + area_re_m1))) 
-gen se2 = 1 /(1+exp(-1*(area_re_s1))) 

-sort area
-by area: gen dup=_n
-by area: egen arearate1=mean(mu1)
-by area: egen arearate2=mean(mu2)

*Here I was able to confirm that arearate1=arearate2.
*I then selected one record per area and calculate the confidence interval below:

-keep if dup==1

gen area_lo=arearate1-1.96*se2
gen area_hi=arearate1+1.96*se2

However, the confidence interval results seem to be incorrect (too wide).  Could someone tell me what I did wrong and what is the correct way to calculate the SE and confidence interval?

Thanks a lot.

Lin
Lin Song
Epidemiologist
Public Health - Seattle & King County
(206) 263-8773



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/