Dear all,

I am using a longitudinal survey with a complex design. Thus, I am using svy command.

- Before starting to perform my estimates with svy, I used to estimate my model as follows:

logit y x1 x2 [pw=weight], cluster(pid)
where pid= personal identifier of the individuals of my sample.


- Now, when I consider the stratification, I set my dataset with svyset in the following way:

svyset psu [pweight=xweight_all], strata(strata) singleunit(scaled)
where psu=the geographical stratum

- and I run my model under the svy prefix:

svy: logit y x1 x2, cluster(pid)

- However, Stata does not allow clustered standard errors. It does not even support xtlogit, which would have been an alternative.
Have you any idea of how to implement clustered standard errors and consider the survey design (other specifications,
avoid svy,...).

Thank you and best,
Lydia