Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: problem with mvprobit


From   "Georg Metzger" <Metzger@zew.de>
To   <statalist@hsphsun2.harvard.edu>
Subject   Re: st: problem with mvprobit
Date   Mon, 19 Nov 2007 15:55:33 +0100

The problem is following:
I want to analyze the effect of experience and have three experience
types, namely no experience, good experience and bad experience. The
good experience variable is the mentioned variable that predicts the
outcome perfectly. I think if I drop this variable it would bias the
result concerning the bad experience measure since the basis category is
not solely "no experience" anymore but "no experience and good
experience", right?

Best, 
Georg


          Georg Metzger
          Zentrum für Europäische Wirtschaftsforschung (ZEW) GmbH
          L 7,1 - 68161 Mannheim
          Telefon: 0621 1235-185
          Telefax: 0621 1235-170
          E-Mail: metzger@zew.de

Zentrum für Europäische Wirtschaftsforschung GmbH (ZEW) 
Centre for European Economic Research 
L 7, 1 · 68161 Mannheim · Germany 
Sitz der Gesellschaft: Mannheim · Amtsgericht Mannheim HRB 6554 
Aufsichtsratsvorsitzender: Finanzminister Gerhard Stratthaus MdL 
Geschaeftsfuehrer: Prof. Dr. Dr. h.c. mult. Wolfgang Franz, Thomas Kohl

------------------------------------------------------------------------


>>> SamL <saml@demog.berkeley.edu> 19.11.07 14:52:56 >>>
It is my understanding that the mvprobit command does not require one
to
have the same exact set of regressors for every equation.  If so, why
not
drop the variable that is predicting perfectly from the equation(s)
for
which it makes perfect predictions, and leave it in the equation(s)
for
which it does *not* make perfect predictions?

HTH
Sam

On Mon, 19 Nov 2007, Georg Metzger wrote:

> Dear all,
> I have a problem in estimating a multivariate probit ('mvprobit'
from
> ssc).
> I want to estimate a multivariate probit model with four equations.
> There is an (important) variable in the specification, which
predicts
> the outcome of the fourth equation perfectly. In a normal probit
> estimation both this variable and the relevant observations would be
> dropped out. However, it is not the case if applying the mvprobit
> command, I think, because the variable and the relevant observations
> contribute to explain the outcome of the other equations. The
mvprobit
> estimation results in outrageous values for both the coefficient and
the
> standard error of this variable. I know that one can't interpret
this
> values; however, does anybody know if including the variable
> significantly bias the other estimation results? Or should I better
> leave out the equation with the perfect fit estimating a trivariate
> specification?
>
> Best,
> Georg
>
>           Georg Metzger
>           Zentrum für Europäische Wirtschaftsforschung (ZEW) GmbH
>           L 7,1 - 68161 Mannheim
>           Telefon: 0621 1235-185
>           Telefax: 0621 1235-170
>           E-Mail: metzger@zew.de 
>
> Zentrum für Europäische Wirtschaftsforschung GmbH (ZEW)
> Centre for European Economic Research
> L 7, 1 · 68161 Mannheim · Germany
> Sitz der Gesellschaft: Mannheim · Amtsgericht Mannheim HRB 6554
> Aufsichtsratsvorsitzender: Finanzminister Gerhard Stratthaus MdL
> Geschaeftsfuehrer: Prof. Dr. Dr. h.c. mult. Wolfgang Franz, Thomas
Kohl
>
>
------------------------------------------------------------------------
>
>
> Zentrum für Europäische Wirtschaftsforschung GmbH (ZEW)
> Centre for European Economic Research
> L 7, 1 · 68161 Mannheim · Germany
> Sitz der Gesellschaft: Mannheim · Amtsgericht Mannheim HRB 6554
> Aufsichtsratsvorsitzender: Finanzminister Gerhard Stratthaus MdL
> Geschaeftsfuehrer: Prof. Dr. Dr. h.c. mult. Wolfgang Franz, Thomas
Kohl
>
------------------------------------------------------------------------
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html 
> *   http://www.stata.com/support/statalist/faq 
> *   http://www.ats.ucla.edu/stat/stata/ 
>

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html 
*   http://www.stata.com/support/statalist/faq 
*   http://www.ats.ucla.edu/stat/stata/

Zentrum für Europäische Wirtschaftsforschung GmbH (ZEW)
Centre for European Economic Research
L 7, 1 · 68161 Mannheim · Germany
Sitz der Gesellschaft: Mannheim · Amtsgericht Mannheim HRB 6554
Aufsichtsratsvorsitzender: Finanzminister Gerhard Stratthaus MdL 
Geschaeftsfuehrer: Prof. Dr. Dr. h.c. mult. Wolfgang Franz, Thomas Kohl 
------------------------------------------------------------------------
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   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