Statalist


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

Re: st: Average Partial Effects after Margeff


From   "C.T. Shehzad" <C.T.Shehzad@rug.nl>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Average Partial Effects after Margeff
Date   Wed, 16 Sep 2009 16:46:50 +0200

Thanks Austin:
The references you gave are interesting. Yes, model is based on theory and has been used before as well. Actually, what puzzles me are the too small standard errors that I get now in addition to, change in significance. May be - margins - command of stata 11 can be a good alternate but I have only stata 10.1.

Thanks and best regards,

Tanveer


Austin Nichols wrote:
C.T. Shehzad <C.T.Shehzad@rug.nl> :
There is no guarantee (in the general case) that any program will give
sensible APE calculations or SEs, including official Stata commands;
you have to check the logic of the program for your particular case,
and in some cases you may have to reparametrize the model or do some
calculations by hand.  See e.g.
http://www.stata.com/statalist/archive/2008-03/msg00652.html
http://www.stata.com/statalist/archive/2009-03/msg01180.html
http://www.stata.com/statalist/archive/2009-07/msg00786.html
http://www.stata.com/help.cgi?margins
and describe your estimation strategy--what are dep,x1,x2,x3?  Are
there logical dependencies among RHS variables?

On Wed, Sep 16, 2009 at 9:31 AM, C.T. Shehzad <C.T.Shehzad@rug.nl> wrote:
Dear Statalisters:

I am using margeff command after xtprobit to get average partial effects.
The problem is that my earlier estimations results (i.e. average partial
effects and their standard errors) from using the previous version of
margeff (2.1.8 ) considerably differ from the current version (2.2.0).
Standard errors are very small now which seem to be a bit unrealistic. I run
the same program file on two different PCs. All other things remain the same
(including actual xtprobit results) but the average partial effects and
especially their standard errors are quite different as calculated in the
two versions of margeff. My program set up is as follows:

xtprobit dep x1 x2 x3, no log
margeff, replace

Moreover, for some variables which are insignificant in xtprobit results
even at 10 percent level, their average partial effects become significant
 at one percent level. Am  I doing something  wrong? or it is just the
updating of the program?

Thanks for help,

Tanveer

Difference in versions:

Version 2.2.0 - 20 August 2009 - margeff reported incorrect standard errors
for marginal effects after oprobit and ologit. This is now fixed.
Version 2.1.9 - 13 August 2009 - margeff reported incorrect average partial
effects after oprobit and ologit. Partial effects calculated at fixed values
or means were however correct.


*
*   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