This Stata webpage nicely describes how to use bootstrap for panel data. However, it is unclear how one can use the svy bootstrap to obtain the distribution of interest when a panel data is used because the svy bootstrap does not allow the cluster or idcluster options to force Stata to select the sample from observations not person-waves. The problem becomes even worse when the mlogit model are used because xtmlogit is not supported by svy prefix. Any ideas?
Thanks,
NM