Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: AW: Probit model with randomly missing obs in the RHS: 2-stage vs joint estimation


From   "Martin Weiss" <martin.weiss1@gmx.de>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: AW: Probit model with randomly missing obs in the RHS: 2-stage vs joint estimation
Date   Thu, 5 Mar 2009 09:23:30 +0100

<> 

Let me guess: Marian uses the default settings in her mail program that
makes her post in rich-text whereas the list only takes text-only messages.
She can change that in her options and post happily ever after. See the FAQs
whose link is attached to every message on the list...

On your prob, I have a hard time believing your first stage. Let me try to
recreate it here


*************
clear*
set obs 1000
*dependent var 70% 0 - 30% 1
g y=runiform()>0.7
*let`s see
prop y
*independent var only observed for dependent equal to 1
g x=rnormal() if y==1
*error message: "outcome does not vary"
probit y x
*************

The fact that x is observed perfectly predicts whether y equals 1. Please
advice if your estimation strategy delivers a different result.


HTH
Martin


-----Ursprüngliche Nachricht-----
Von: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von Tom Trikalinos
Gesendet: Donnerstag, 5. März 2009 01:44
An: statalist@hsphsun2.harvard.edu
Cc: abi.dsb99@gmail.com
Betreff: st: Probit model with randomly missing obs in the RHS: 2-stage vs
joint estimation

Sending on behalf of Marian, who has problems posting to the list --
forgive the title, it is mine, not Marian's (-- and i do not claim i
comprehend the question.)
t

-----------------------------------------------------
Marian writes:

Hi,
I need to estimate a probit with a continous variable in the RHS that
it is only observed when the binary outcome is 1. I first  estimate
the continuous variable with those observations where there is no
censoring and estimate predicted values for all the observations.
(Selection is random)
Then I set the predicted value for all observations in the RHS to
estimate the probit and correct standard errors  (Topel-Murphy) for
the fact I include an estimated variable in the model.
How may I estimate both equations jointly (or recursively) and
efficiently instead of in two-stages?
Thank you,
Marian

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index