Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: svyset with raked weights


From   Steven Samuels <[email protected]>
To   [email protected]
Subject   Re: st: svyset with raked weights
Date   Wed, 9 Mar 2011 15:29:05 -0500

Fernando wrote to me privately:

> 
> Steve,
> Thank you very much for your comments.
> In fact, before starting this thread, I was thinking in using the raked (or calibrated) weights as pweights, but I hesitated after reading one of your previous posts to statalist (http://www.stata.com/statalist/archive/2009-09/msg00891.html), where you say: "Note that if  your colleagues treat the post-stratification weights as ordinary pweights, they might not get the same standard errors as Stata does".
> Now I think I've caught the idea of not going further with the sample.
> 
My 2009 statement was correct.

Specifying post-stratification weights  and strata in -svyset- does lead properly to slightly different  standard errors (compared to design stratification) because the  the sample sizes in the post-strata were not fixed by the design.  The use of raked weights in a [pweight=] statement  does not lead to different standard errors  unless a replication method is used to estimate standard errors. (Then the weights can be re-estimated for each replicate. See the -help- for Stas Kolenikov's contributed program -bsweights- (-findit-).) 

However the effect of ignoring the raking is likely to be small, compared to other sources of error, unless than the post-strata sample sizes are too small. The danger, especially with smaller post-strata sample sizes, is that raking will increase standard errors because it increases the CV of the weights.  


Steve


Steven J. Samuels
Consultant in Statistics
18 Cantine's Island
Saugerties, NY 12477 USA
Voice: 845-246-0774
Fax:   206-202-4783 
[email protected]




El 08/03/2011 21:56, Steven Samuels escribió:
> 
> Fernando-
> 
> Simply reference the raked-weight variable in the [pweight=] statement.
> 
> 
> Steven Samuels
> 
> 
> 
> 
> On Mar 8, 2011, at 2:46 PM, Fernando Terrés wrote:
> 
> Steve,
> Thank you, I didn’t know about GREG in Stata (is it more appropriate to use raked weights, as the data is mainly categorical?). My question goes one step further, which is the correct svyset design for inferences, given that raked (or calibrated) weights are available.
> 
> 

-- 
Fernando Terres
Lecturer. PhD candidate
CERPIE/Research
Universitat Politecnica de Catalunya
Diagonal, 647 - 10
08028 Barcelona (Spain)
+34.934.010.708
[email protected]
http://cerpie.upc.edu


*
*   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index