RE: st: question on survival analysis and selection correction

["Craig, Benjamin M." <Benjamin.Craig@moffitt.org>]

I respectfully disagree. Control function techniques such as Heckman's require the assumption of the correlation function(i.e., control function), not the full distribution. Only the heckman with MLE requires the full distributional assumptions (which is why it often produces smaller p-values). There are loads of IV approaches that do not require distributional assumptions. Nevertheless, the assumptions that they require may be equally erroneous, particularly those involving instrument strength. 

Joseph Terza has developed an iv estimator for survival models, so you may wish to contact him directly. He is very helpful. http://www.uncg.edu/bae/people/terza/

Alternatively, counts of days or years can be modeled with an iv count models like iv poisson or neg bin. This would be the easiest answer (but loaded with parametric assumptions). See the work of John Mullahy at Wisconsin.

Cheers,
Ben



 


Benjamin M. Craig, PhD

Assistant Member, Health Outcomes & Behavior, Moffitt Cancer Center

Associate Professor, Economics, University of South Florida 

12902 Magnolia Dr, MRC-CANCONT, Tampa, FL 33612-9416

Phone (813) 745-6710; Fax (813) 745-6525 

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Maarten buis
Sent: Tuesday, September 28, 2010 5:17 AM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: question on survival analysis and selection correction

--- On Tue, 28/9/10, Gáti Annamária wrote:
> I am doing survival analysis on a graduate dataset.
> I analyze "how long graduates survive in unemployment" after graduation.
> Some find a job in 1 month, some in 3 etc, some not even in 30 month.
>
> There are graduates in my data who do not search for a job after 
> graduation as they already had one while doing their studies (they 
> mostly did evening courses etc.).
>
> I suppose there is a selection bias I should account for while I am 
> doing my survival analysis as those who search for a job after 
> graduation are mostly young, doing their studies full time and 
> therefore they are definitely different from the ones who earn their 
> degrees while working.
>
> Is it possible to use the Heckmann correction methods together with 
> survival analysis? That is the only method I have learnt a bit about.
> If I need to do this correction what is the way to specify it in my syntax?

The whole Heckman correction stuff critically depends on the assumption that the residuals the selection and main equations folow a multivariate normal distribution. With survival analysis, you cannot assume normally distributed errors, so that cannot work.

I would just consider them two completely different populations, and just study the full time student population. In the continental west European context that makes perfect sense: these groups face a completely different labor market, and are in no way competitors, they also hardly meet or interact during their studies. 

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------


      

*
*   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/


This transmission may be confidential or protected from disclosure and is only for review and use by the intended recipient. Access by anyone else is unauthorized. Any unauthorized reader is hereby notified that any review, use, dissemination, disclosure or copying of this information, or any act or omission taken in reliance on it, is prohibited and may be unlawful. If you received this transmission in error, please notify the sender immediately. Thank you.
*
*   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/