<>
Martin helpfully provided
qui reg pr we le c.we#c.le
margins, dydx(*)
Note that the results of that more concise calculation are exactly in
agreement with the -lincom- method I posted, with one exception: -
margins- constructs confidence intervals based on the Normal
distribution and z-scores, while -lincom- does so with the t
distribution (the difference that the -small- option invokes on
several estimation commands). Naturally for a sample of reasonable
size this matters not for inference.
You can also do things like
margins, dydx(*) at(weight=4000)
margins, dydx(*) at(weight=4000 length=200)
to see how these derivatives change over the regressor space.
Kit Baum
[email protected]
*
* 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/
