Re: st: Binary Choice Panel Data Model with Sample Selection.

[Ippab <ippab.statalist@gmail.com>]

Hi Melissa,

Thanks for your excellent suggestions.  I used the leads you provided
and was able to estimate the heckprob model with bootstrapped errors.
The only thing is that mfx (stata 10.1) does not take vce() as an
option.

By the way, what are the major limitations of following this approach?
 I suspect the clustered bootstrapped errors are reasonable substitute
of random effects, is that correct?

Best,

ippab

On Fri, Feb 11, 2011 at 8:28 AM, Garrido, Melissa
<melissa.garrido@mssm.edu> wrote:
> Hi Ippab -
>
> See http://www.stata.com/support/faqs/stat/xt_boot.html  for a suggestion on dealing with standard errors in panel data.  You can use the vce option with heckprob.
>
> Also, have you played around with your model at all to see what is causing your model not to converge?  Maarten Buis gives some good examples of what to look for in: http://www.stata.com/statalist/archive/2010-08/msg00847.html. That discussion refers to an ivprobit model, but the same issues are relevant here.
>
> Assuming you want the marginal effect of explanatory variables on your outcome variable, given a value of 1 for the selection equation, I believe you can use:
> margins, predict (pcond) dydx(*) vce()
> where you would include the same options for vce that you do in the heckprob equation. I'm not completely certain on this last point - hopefully others have suggestions for you as well.
>
> Melissa
>
>
>
> --------------------------------------------------------------------
> Melissa Garrido, PhD
>
> Research Health Science Specialist
> GRECC/REAP, James J Peters VA Medical Center
> 130 West Kingsbridge Road
> Bronx, NY 10468
> 718-584-9000 x 3804
> melissa.garrido@va.gov
>
> Assistant Professor
> Brookdale Department of Geriatrics & Palliative Medicine
> Box 1070
> Mount Sinai School of Medicine
> One Gustave L. Levy Place
> New York, NY 10029
> melissa.garrido@mssm.edu
>
>
> ----------------------------------------------------------------------------
>
> From      Ippab <ippab.statalist@gmail.com>
> To        statalist@hsphsun2.harvard.edu, Stata Technical Support <tech-support@stata.com>
> Subject           st: Binary Choice Panel Data Model with Sample Selection.
> Date      Thu, 10 Feb 2011 18:23:43 -0700
>
> Hello Stata Users,
>
> I have a panel data with censored observations (sample selection),
> where the outcome variable is binary.  Therefore, I need to estimate
> both the selection equation and outcome equation with binary choice
> models.  I am wondering if there is any command (user written or stata
> provided) for estimating such a heckman model with panel data (I am
> interested in both fixed and random effects).
>
> I have tried to use the heckprob command, but it does not converge.
> Also, my understanding is that it does not consider the panel nature
> of the data.  I was able to find a reasonable instrument for the
> selection equation, but it still did not work.
>
> Is it reasonable to do the estimation along the line of two step
> heckit approach?  In particular, if I get the mills ratio from a
> probit model in the first step and include that as a control variable
> in the xtlogit or xtprobit with the censored data, would it be
> correct?  I followed this approach and the mills ratio is significant.
>  From reading the relevant literature, even if this is correct, I need
> to correct the standard error in implementing the two step method.
> How do I correct the standard errors after running xtlogit or
> xtprobit?  Finally, how do I calculate the conditional marginal
> effect?
>
> I would really appreciate any ideas, suggestions, or directions you
> may have.  Thanks a lot in advance!
>
> Best regards,
>
> ippab
>
> *
> *   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/
>

*
*   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/