Re: st: -xtmixed- performance problem

[rgutierrez@stata.com (Roberto G. Gutierrez, StataCorp)]

Hobst <tobias.friedli@access.uzh.ch> writes:

> I want to run the following two-way error component regression in xtmixed,
> with cluster robust standard errors

> xtmixed y x1 x2 x3 x4 || _all: R.id || datevar:, mle residuals(, by(id))

> Without the term residuals(,by(id)) the regression is very fast, but with
> the the option included, it did only 2 iterations in 48h.

> Does anybody have any idea, what i could do to speed the regression up? Is
> there maybe an option to decrease accuracy or something like that?

In a later post on this thread you said you have 320 groups each with 12
observations.  When you added -residuals(, by(id))- in the above you specified
a heteroskedasticity model where you estimate a distinct residual variance for
each group.  This added 320 parameters to be estimated in your model, hence
the slowdown.

Cluster robust standard errors are currently not supported by -xtmixed-,
although that is something we are looking into adding in the future.

--Bobby
rgutierrez@stata.com
*
*   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/