Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: svy bootstrap and nlsur function evaluator program


From   Stas Kolenikov <skolenik@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: svy bootstrap and nlsur function evaluator program
Date   Mon, 6 Dec 2010 13:42:12 -0600

On Mon, Dec 6, 2010 at 1:32 PM, Stefano Verde <verdes@tcd.ie> wrote:
> Dear Stas,
>
> I have purchased and read your paper ("Resampling variance estimation
> for complex survey data"), and also looked at other documents.
> However, I'm still not sure whether -bsweights- really is what I need.
>
> The thing is that my dataset only includes information on the gross
> factor (i.e. the sampling weights), which are supposed to embody
> already the choices on survey design, such as stratification,
> clustering and non-response. Thus, information about the latter is
> just not there.
>
> So, do I still need to create replicate weights using -bsweights-?
>
> Isn't there a way of bootstrapping the elasticities from my model
> (complex functions of estimated parameters) just using the gross
> factors (i.e. sampling weights)?

The weights can only reflect the probabilities of selection (and
sometimes also contain non-response and post-stratification
adjustments). They don't have anything to say about clustering or
stratification. (One of the homework assignments I had in the sampling
class was to come up with two different designs that would produce the
same point estimates, but different standard errors for a given
sample. In fact, of course, there are infinitely many such designs,
and you can get standard errors that can range anywhere from zero to
more than the sqrt of [sample variance/n], or may even be undefined,
as is the case of systematic sampling). You would need to go back to
the data providers and check with them what the sampling design really
was.

You may indeed have been lucky, and only needed the weights if the
design is close to SRS (except for varying probabilities of
selection). Some nationally representative phone surveys come pretty
close to that ideal.

-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
*   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/


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