"repeated time values within panel" and "convergence not achieved"

[Adis Dan]

I am Adisalem from Ethiopia. I glad for joining this platform

I was trying to fit Generalized Estimating Equation model (GEE) using one of the DHS data set. I faced difficulty in the following issues

1. setting the panel data for GEE model i.e. xtset command didn't work for me. When I type the command it replied that there is "repeated time values within the panel". I tried to diagnose the problem and identified duplicate caseid. But failed to drop these duplicate caseid because of the power of the study. Would you suggest me any alternatives to handle this case without dropping observations?

2. The second one was for one of my dependent var the iteration worked but under the Stata output for exchangeable correlation structure it says "convergence not Achieved". I tried with different working correlation structure and; with the independent one I didn't actually get this convergence message.But for the rest of the correlation structure the command replies saying repeated time values within panel?

would you please help me address this problem?

What should be considered while selecting the appropriate correlation structure?

thank you so much

[Andrew Musau]

1. setting the panel data for GEE model i.e. xtset command didn't work for me. When I type the command it replied that there is "repeated time values within the panel". I tried to diagnose the problem and identified duplicate caseid. But failed to drop these duplicate caseid because of the power of the study.

If these are perfect duplicates, i.e., the entries of all variables are the same for a unit in the dataset, then you do not increase the power of your study by keeping the duplicate observations. Start with

duplicates tag *, g(dup)
sort caseid
browse if dup

If you see duplicates, you should drop them and then check whether the "repeated time values" error disappears. If it does not, there is something with your data that you are not understanding. Perhaps the unit of observation is not "caseid". You will need to provide a sample of the data using dataex to push this forward (see details in FAQ Advice #12).

[Adis Dan]

Thank you Andrew for your kind response
I think it is better to post sample data here for better understanding