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

Re: AW: st: RE: wald tests with mfx

From   Maarten buis <>
Subject   Re: AW: st: RE: wald tests with mfx
Date   Tue, 14 Jul 2009 21:44:39 +0000 (GMT)

--- On Tue, 14/7/09, Rich Steinberg wrote:
> No, we are comparing like with like.  So  when we evaluate 
> our typical person who received an inheritance, we evaluate
> how earned income, inherited income, etc. affect giving by
> people at the mean of the subsample receiving an inheritance. 
> It is only within this group that we test equality of the
> coefficient on earned income and inherited income. 

sounds reasonable

> I still haven't gotten a response on an elegant way to do
> what we need in stata, where we use a single tobit to
> produce all these subsample marginal effects and do the
> appropriate within equality tests, and the problem is that
> all the postestimation stuff refers to the tobit, and not
> the changes on that tobit resulting from our mfx
> calculation.  

If you look at what -mfx- returns you will see that it only
returns the standard errors of the marginal effects, not
the entire variance covariance matrix of these marginal
effects. To do these equality tests you also need these
covariances, so you can't use -mfx- for this purpose.

What you could do is derive / look up in a textbook the
formula of these marginal effects and compute them
using -nlcom-, this will also compute a variance covariance
matrix, so will allow you to do the equality tests.

Hope this helps,

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen


*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index