--- On Thu, 2/7/09, [email protected] wrote:
> I am running these two types of commands in stata:
>
> bootstrap, reps(200): xtgee depvar indvar, fam(normal)
> link(id) i(case_no) t(week1_10) robust
>
> xi: xtgee depvar i.indvar, fam(normal) link(id) i(case_no)
> t(week1_10) vce(bootstrap, reps(200))
>
> The two commands give similar results for some of the
> variables, but for some of the variables I get very
> different results with respect to significance, for most of
> the variables I get significant results with the 1st command
> given above.
There are two answers to this:
First the bootstrap is a (pseudo) random proces, so you can
expect to get different estimates of the standard error if
you don't set the seed, see for example the example below:
*--------- begin example -------------
sysuse auto, clear
qui bootstrap , reps(200): reg price mpg
est store noseed_a
qui reg price mpg, vce(bootstrap, reps(200))
est store noseed_b
set seed 12345
qui bootstrap , reps(200): reg price mpg
est store seed_a
set seed 12345
qui reg price mpg, vce(bootstrap, reps(200))
est store seed_b
est table _all, se
*---------- end example --------------
If this was the only problem than these differnces
in results would indicate that 200 replication
isn't enough in your case to get stable estimates,
and that you would need to increase that.
However, there is also a second issue with your
sequence of commands, and that is that -bootstrap-
by default does not respect the nested nature of
your data, while the -vce(bootstrap)- option
automatically takes that into account.
Hope this helps,
Maarten
-----------------------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
*
* 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/
