Re: st: Calculating margins after biprobit

[Tirthankar Chakravarty <tirthankar.chakravarty@gmail.com>]

I corrected you based on what you sent to the list. As in the example
I sent, you need to enter your model in Stata with the included
exogenous variables in the main/"second stage" regression and all
exogenous variables in the "first stage" regression.

T

On Tue, Dec 6, 2011 at 9:47 AM, Wakeman, Simon <Simon.Wakeman@esmt.org> wrote:
> Thanks both for your comments.
>
> Tirthankar Chakravarty: I am not clear as to what is your criticism of the way I have estimated the binary outcome. Unfortunately I do not have the 2nd edition of Wooldridge's text (only the first) but from what I can gather from your example the difference is that you include Z on the RHS of both first and second equations (which is actually what I had done - I had just mistakenly omitted it). Please can you confirm.
>
> I am interested in precisely the marginal/partial effect you mention (P[Y1i=1|Y2i=1, Zi] - P[Y1i=1|Y2i=0, Zi]) so this is very helpful.
>
> urbain thierry YOGO: I still get the error message when I try the command you suggested. I can overcome this by adding "force" to the options, but it does not solve the issue of whether I have a potential problem.
>
> Simon
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Tirthankar Chakravarty
> Sent: Tuesday, December 06, 2011 12:10 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: Calculating margins after biprobit
>
> Two comments:
>
> 1) The way that you have estimated the binary outcome with binary endogenous variables model using -biprobit- is not correct. Here is a replication of an example in Wooldridge, "Econometric Analysis of Cross-Section and Panel Data", 2nd ed., 2010, MIT Press, to be found in table 15.2, pg. 598:
>
> *********************************************************************************
> use http://fmwww.bc.edu/ec-p/data/wooldridge2k/LABSUP, clear biprobit (worked=nonmomi educ c.age##c.age black hispan morekids) ///
>        (morekids=nonmomi educ c.age##c.age black hispan samesex)
> *********************************************************************************
>
> Here "morekids" is the binary endogenous variable and "worked" is the binary outcome. Note how the model is entered in Stata. The coefficient on the binary endogenous variable is -0.703, as reported in the text.
>
> 2) It is important to be sure what you mean by marginal/partial effects in this context. One such quantity of interest, which can be called partial effect of a change in the binary endogenous variable, for an individual would be:
>
> P[Y1i=1|Y2i=1, Zi] - P[Y1i=1|Y2i=0, Zi]
>
> where Y1 is the binary outcome and Y2 is the binary endogenous variable; Z are all the exogenous variables in the system. This can be constructed in Stata as the nonlinear prediction:
>
> ***********************************
> predictnl probch = predict(pcond1) - ///
>   predict(p10)/(1-predict(pmarg2))
> ***********************************
>
> after the estimation above. The sample average of this quantity (use
> -summ-) is the average partial effect, where partial effect is defined as above. Other "partial effects" of interest can be defined.
>
> Please see section 15.7.3 in Prof. Woodridge's book for an excellent discussion of these issues.
>
> T
>
> On Mon, Dec 5, 2011 at 12:37 PM, urbain thierry YOGO <yogout@gmail.com> wrote:
>> Try the following command
>> margins, dydx(_all) post
>> this is equivalent to  mfx compute, predict(p11) with
>> (p11)=(depvar1=1, depvar2=1) however a simple and general way is the
>> following
>>
>> The marginal effects for, Pr(depvar1=1, depvar2=1), are mfx compute,
>> predict(p11) The marginal effects for Pr(depvar1=1, depvar2=0) are mfx
>> compute, predict(p10) The marginal effects for Pr(depvar1=0,
>> depvar2=1) are mfx compute, predict(p01) The marginal effects for
>> Pr(depvar1=0, depvar2=0) are mfx compute, predict(p00)
>>
>>
>>
>> 2011/12/5, Wakeman, Simon <Simon.Wakeman@esmt.org>:
>>> I am investigating the effect of a binary variable X on a binary variable Y,
>>> where X is potentially endogenous. I estimated a biprobit:
>>>  biprobit (Y = X ) (X = Z).
>>>
>>> However, when I use the following margins command
>>>  margins, dydx(X) over(X)
>>> I get the error message:
>>> "default prediction is a function of possibly stochastic quantities other
>>> than e(b)"
>>>
>>> I also tried mfx but it does not seem to produce a marginal effect for X.
>>>
>>> I would much appreciate if anyone can tell me how to obtain the marginal
>>> effect of X on Y.
>>>
>>> Thanks for your help.
>>>
>>> Simon
>>>
>>>
>>> *
>>> *   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/
>>>
>>
>>
>> --
>> *Urbain Thierry YOGO
>> P.h.D candidate in Economics*
>> *
>> *   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/
>
>
>
> --
> Tirthankar Chakravarty
> tchakravarty@ucsd.edu
> tirthankar.chakravarty@gmail.com
>
> *
> *   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/



-- 
Tirthankar Chakravarty
tchakravarty@ucsd.edu
tirthankar.chakravarty@gmail.com

*
*   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/