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: command for penalized MLE using a complex survey data?


From   "Kim, Isok" <[email protected]>
To   "'[email protected]'" <[email protected]>
Subject   RE: st: command for penalized MLE using a complex survey data?
Date   Fri, 28 Sep 2012 10:24:18 -0400

Thanks Stas!  It sounds like penalized MLE using -firthlogit- can't be applied to survey sample - at least from the theoretical perspective. Would it be correct to assume this?

isok

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Stas Kolenikov
Sent: Thursday, September 27, 2012 8:09 PM
To: [email protected]
Subject: Re: st: command for penalized MLE using a complex survey data?

Firth's correction deals with the information matrix in the i.i.d.
likelihood context. The i.i.d. assumption is explicitly violated with the -svy-type data. Besides, the target of inference with survey data are finite population parameters, which in this case would be the parameters of the census logistic regression. It is difficult to say how one could correct the finite sample bias with respect to that parameter. I am not sure as to whether one can even find unbiased estimating equations in this situation.

That's all theory, of course, and when the referees ask for this or that, they may not know that such a thing does not exist. What I did for one of my projects was to run -firthlogit- with [aw], which I think it sort of does, and then -ereturn repost-ed the variance-covariance matrix from -svy: logit-. This has no justification whatsoever, of course, but that's arguably the closest you can get.

--
-- Stas Kolenikov, PhD, PStat (SSC)  ::  http://stas.kolenikov.name
-- Senior Survey Statistician, Abt SRBI  ::  work email kolenikovs at srbi dot com
-- Opinions stated in this email are mine only, and do not reflect the position of my employer


On Thu, Sep 27, 2012 at 5:12 PM, Kim, Isok <[email protected]> wrote:
> Hi,
> I noticed that firthlogit is not supported by svy prefix command with vce(linearized).  Does anyone know other commands to run penalized maximum likelihood estimation using the survey data?
>
> Thanks!
> isok
>
> Isok Kim, PhD LCSW
> Assistant Professor
> School of Social Work
> University at Buffalo
> 667 Baldy Hall
> Buffalo, NY  14260-1050
> (716) 645-1252
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


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