 Bootstrap sampling and estimation in the Stata documentation

Official StataCorp manual, bootstrap — Bootstrap sampling and estimation. Printed pp. 3–5, Options, specify that cluster() samples clusters, that unequal cluster sizes give varying replicate row counts, and that idcluster() assigns a distinct identifier to each resampled cluster copy. Printed p. 18, Technical note, explains why resampled panel identifiers must distinguish repeated copies. This source supports the resampling algorithm; it does not promise valid inference for every clustered-data model.