One thing you can try is to look into Nick Winter's -svr- code to see
how he created jackknife weights.
If I were to do this, I would take yat another look at Rao & Wu paper
on resampling with survey data (Journal of the American Statistical
Association, 83:231-- 241, 1988), figure out the correct number of
clusters to get resampled, check for other caveats, and then, in
Stata-like pseudocode,
g double bsweight = 0 if ~mi( actual_pweight )
foreach `h' in the list of strata {
forvalues k=1/desired number of resampled clusters {
local random_cluster = uniform()*(# clusters in the current stratum) + 1
replace bsweight = bsweight + actual_pweight if stratum == `h' &
cluster == `random_cluster'
}
}
This should give you one bootstrap resample. Now, the question
remains, at least to me, what one is to do with the 0 weights (should
they indeed be zero or missing? I would keep them at zero), and what
happens if in a certain stratum, only a single cluster has been
randomly selected (I would probably discard such subsample). See if
that is at all addressed in the paper above, and try to find some
follow-ups -- and please let us know if you are successful.
Hope Nick Winter or Jeff Pitblado would comment on the above. It would
be nice if -svy- were to take in the bootstrap weights, too -- I don't
think it does currently. You can probably get around by feeding them
in as -jackknife- weights to the aforementioned -svr-.
On 4/14/06, Dongyi Du <[email protected]> wrote:
> hi, all:
>
> I have information of weight, psu, and strata for a survey data, can
> anyone tell me if I can generate my own bootstrap weights? if I can,
> how?
>
> thanks.
>
> Dong
>
> *
> * 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/
>
--
Stas Kolenikov
http://stas.kolenikov.name
*
* 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/
