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: Obtaining population estimates for multiply imputed survey data


From   Steve Samuels <[email protected]>
To   [email protected]
Subject   Re: st: Obtaining population estimates for multiply imputed survey data
Date   Mon, 14 May 2012 11:35:14 -0400

Good points, Richard.

I should have mentioned in my post that, independent of -mi estimate-, the validity of -svy: total- depends on the same conditions I mentioned for utilizing the sum of the weights as an independent estimate of population size. Elaborating slightly, the conditions are:

1) that weights be corrected for non-response or that non-response be very low
2) that the weights not be standardized to sum to sample size (a regrettable practice in my opinion)
3) that the design be appropriate: PPS should be OK.


Steve
[email protected]


On May 14, Richard Williams wrote


I don't know when it [-cmdok-] got added, but it is an option in Stata 11.2, which the original poster should be able to upgrade to.

My only hesitation in using -cmdok- with -total- is that I wonder if Stata had some good reason for not building in support to begin with. Whenever Stata surprises me, I am never sure if it is because (a) Stata is smarter than I am, or (b) somebody just hasn't gotten around to programming something yet.



On May 14, Steve Samuels wrote:

Also, -cmdok- was not an option in Stata 11. If it is not in Stata 11.1, Lauren
can -mi estimate- means and standard errors,then multiply these by an estimate
of population size N. If a good external estimate of N is available, I'd
recommend that. Also for many surveys an independent estimate is available: the the sum of the sample weights, corrected for non-response. This won't work if: 1) the weights were standardized to sum to sample size; or 2) the sample was drawn poorly, e.g. a SRS of unequally sized units was drawn at the first stage.

Steve
[email protected]




At 09:09 PM 5/13/2012, Stas Kolenikov wrote:
I'd say this is an oversight. There's nothing special to -total- that
would make it terribly different from other commands, so you should
just proceed with -cmdok- option of -mi estimate-.

Support for -total- was added in version 12. It does seem odd it wasn't supported before. Perhaps it could be double-checked against the -means- command to make sure it will work ok in Stata 11.


On Sun, May 13, 2012 at 11:24 AM, Lauren Pollack
<[email protected]> wrote:
I am using Stata 11.1 and trying to obtain population estimates for
multiply imputed survey data. My data is mi set and I am analyzing it
using the mi commands. It is survey data, so it is also svyset. I can
see that mi commands allow one to obtain proportions, ratios, and
means for survey data using the command structure "mi estimate: svy:
[proportion/mean/ratio]" but the command I really need to use is "svy:
total", to obtain population estimates. This is the command I use to
obtain population estimates for non-mi data. Is there any way to do
this using the mi commands? It's not listed as a supported command,
but I have to imagine there is some way of obtaining population
estimates for multiply imputed survey data.

*
*
*   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