Re: st: mi Impute with survey data

[Stas Kolenikov <skolenik@gmail.com>]

On Thu, May 3, 2012 at 9:17 AM, Fernando Rios Avila <f.rios.a@gmail.com> wrote:
> I ve been reviewing the information available respect to Multiple
> imputation from the Stata manual 12, and one of the first
> recommendations when building the model for the imputation is to
> include the survey structure in the modeling, with information such as
> Weights, strata and clusters. However, i dont seem to find any other
> reference in the -mi impute- section that explains how to include
> strata in the modeling.
> What should be the correct way to include strata in the imputation model?

Put strata dummies into every equation. Note that by using the
imputation model that uses information across strata, you lose the
great advantage of stratified samples, namely, that the estimates
based on different strata are independent. Hence, the -strata()-
option of -svyset- is no longer appropriate.

My understanding is that the interplay of multiple imputation and
thorough design-based survey statistics is still underdeveloped. I am
trying to keep an eye on both, but haven't seen anything decent since
Shao & Sitter (1996,
http://www.citeulike.org/user/ctacmo/article/1269394).

-- 
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/