Sometimes observations on the outcome variable are independent across groups (clusters), but are not necessarily independent within groups. Many Stata estimation commands support the cluster() option that allows you to specify a variable that indicates which group each observation belongs to. For example the variable famid might represent which family one belongs to and the cluster(famid) option indicate that observations might be correlated within (not independent) within families.
More text on this forthcoming
- Stata help and manuals
- More on cluster option in Stata help for estimation options (also [U] 20.14 Obtaining robust variance estimates)