Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: xtreg and vce(bootstrap)


From   jpitblado@stata.com (Jeff Pitblado, StataCorp LP)
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: xtreg and vce(bootstrap)
Date   Wed, 01 Nov 2006 14:09:35 -0600

Jennifer Alix-Garcia <Jennifer.Alix-Garcia@mso.umt.edu> asked about the
resampling scheme used by -vce(bootstrap)- when used with -xt- commands:

> It isn't clear to me what kind of sampling method the vce(bootstrap)
> command uses when you specify it as an option for xtreg with fixed
> effects with a panel.  Does it use a "block bootstrap" method, i.e.,
> drawing strings of time series from within the i() variable?  If so, how
> long of a block does it draw? If not, how is it dealing with serial
> correlation?  Thanks in advance for your insights.  Best, Jennifer

Austin Nichols <austinnichols@gmail.com> replied that the panels are resampled
just like clusters:

> I don't have a manual handy, but I believe the approach is to sample
> panels (thus, strings of time series corresponding to the whole panel,
> in your parlance) with replacement, much as the cluster option on
> -bootstrap- does.  Also note
> (from update 06oct2006)
> 1. areg with option vce(bootstrap) or vce(jackknife) resampled
> observations instead of the groups identified by the absorb()
> variable. This has been fixed.

Jennifer then wrote in that she tried using the -bootstrap- prefix command
with the -cluster()- option but got different results:

> Hmmm...I did try bootstrapping it using the bootstrap cluster() option
> and got somewhat different results, which was on reason why I was
> wondering about the sampling used for vce(bootstrap).  I will try to get
> my hands on a manual and figure out if it specifies the sampling
> strategy.  Thanks for your help.

Jennifer likely didn't use the -idcluster()- option for her comparison with
the -vce(bootstrap)- results.  For example,

	. webuse nlswork
	. xtreg ln_w grade age*, i(idcode) fe vce(bootstrap, seed(1234))

is equivalent to

	. bootstrap, seed(1234) cluster(idcode) idcluster(newid) :
		xtreg ln_w grade age*, i(newid) fe

(the above being one line, but broken up here for aesthetics)

Notice that -bootstrap- generates the resampled datasets by clustering on
-idcode-.  The -idcluster()- option causes the -newid- variable to be
generated (for each new bootstrap sample dataset), whose values identify the
resampled clusters uniquely; and -xtreg- uses these resampled panels
via the -i(newid)- option.

--Jeff
jpitblado@stata.com
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index