st: Bootstrapping xtmixed

["Juan Dominguez (Med)" <juan.dominguez@monash.edu>]

Dear Statalististers,

I have a question relating to the specification of clustervar and
panelvar when bootstrapping xtmixed. According to the bootstraping
documentation regarding longitudinal data,

"To ensure that the combination of (clustervar, panelvar) values are
unique in each of the bootstrap samples, cluster(clustervar),
idcluster(newclustervar), and group(panelvar) must be specified with
bootstrap, and i(panelvar) must be used with the main command."

However, xtmixed does not allow the i(panelvar) option (as I found out
when trying to run it). By the way, xtmixed does not have/allow the
vce-options available for simpler xt procedures either.

I am fitting a group x time covariate random slopes model (with three
levels in group and two levels in time) with the following command:

. xtmixed var group##time covar1 covar2 || id:time

I have used bootstrap in the following two ways with results that are
consistent with the non-bootstrapped model:

. bootstrap, cluster(id): xtmixed var group##time covar1 covar2 || id:time

. bootstrap: xtmixed var group##time covar1 covar2 || id:time

Do I still need to find a way of specifying both the clustervar and
the panelvar? If so, how do I deal with the fact that xtmixed does not
allow the i(panelvar) option? Or is bootstrap: xtmixed able to figure
out which is which without specification as in the above examples?

Any help would be greatly appreciated.

Juan
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/