--- On Sat, 13/2/10, Xi Pan wrote:
> I am using Stata 9.2.
> I run mlogit model and then use mfx to obtain the marginal
> effect for each category. But covariates consist of ‘years
> of work experience’, its squared term and other variables
> such as gender.<snip>
> 1. How to calculate the standard error for this true
> marginal effect for experience?
I have four thoughts about that.
1) You have to do that using -nlcom-.
2) If you switch to Stata 11, you can use the -margins- command to
get those directly.
3) In non-linear models like -mlogit- there is no such thing as "the
true marginal effect". Think for instance about the difference the "marginal effects at mean values of the explanatory variables" and
the "mean marginal effect". some argue that the latter is the true
marginal effect, but in fact both are plagued with too many
problems to deserve that title.
4) There is an inconsitency with your request: You added the square
term to make the effect of that variable non-linear, and now you
want to summerize its effect with one number, i.e. find some linear
effect that "best" summarizes the non-linear effect. If that is
your only aim, then the logical choice would be to remove the square
term...
Hope this helps,
Maarten
--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany
http://www.maartenbuis.nl
--------------------------
*
* 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/
