Statalist The Stata Listserver


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

st: Average partial effects after oprobit using margeff


From   "Simo Goshev" <goshevsg@mcmaster.ca>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: Average partial effects after oprobit using margeff
Date   Wed, 9 May 2007 07:57:30 -0400

Dear Stata users,

I have an issue with the estimation of average partial effects after
-oprobit- using -margeff-. Here is what I am observing: 

I run:

oprobit y x1 x2 x3
margeff

Then, I calculate manually the effect of a change in x1 (which is
continuous) in the first outcome category using the following steps:

oprobit y x1 x2 x3 
predict xb if e(sample), xb
gen out1 = (-1*(_b[x1])) * (normalden(_b[/cut1]- xb))
sum out1

The calculated mean is the average partial effect of x1.
Now, when I compare the estimated APE from -margeff- with the above mean,
the manually estimated effect turns out to be more than two times bigger in
magnitude than the one provided by -margeff-. 

I also changed the sample and repeated the procedure. The effects were still
substantially different (although this time the -margeff- provided effect
was larger).

I would appreciate any ideas or suggestions.

Best regards,
Simo Goshev


*
*   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