# st: Re: mfx, molasses, and outreg

 From Christopher Baum To statalist@hsphsun2.harvard.edu Subject st: Re: mfx, molasses, and outreg Date Wed, 15 Jan 2003 06:57:41 -0500

On Wednesday, Jan 15, 2003, at 02:33 US/Eastern, statalist-digest wrote:


My second regression is a tobit model with robust standard errors.

From James Hardin's FAQ "How can I get robust standard errors for tobit"
I learn that I should use the intreg command. This is fine, but then, in
order to obtain unconditional marginal effects I have to use the <mfx
compute> command. This causes two major problems: (i) mfx compute is
possibly the slowest thing I have ever met in Stata. In fact, as of
today, I have been unable to wait long enough for mfx to produce some
output (admittedly, I may not be patient enough...), and (ii) even if
willing to wait for mfx compute to stop torturing my hard drive, outreg
seems unable to make use of mfx output (please tell me I'm wrong).

My question: how can I bypass the use of mfx compute, and get intreg-
marginal effects available to outreg?
If you don't need standard errors for what mfx computes, use the 'nose' option, which makes it run much faster.

Here is a code fragment in which I add statistics generated by mfx to regression output (from newey) that I would like outreg to display:

newey2 ratb'_sd Lwcvgdp infl rleadinc,lag(4)
estimates hold col2,copy
mfx compute, eyex
mat nu = e(Xmfx_eyex)
local eta = nu[1,1]
mat nuse = e(Xmfx_se_eyex)
local etase = nuse[1,1]
estimates unhold col2
outreg using gtabletable', merge se 3aster bracket /*
*/ addstat("$\hat{\eta}$",eta', "s.e.",etase')

Note that one has to be careful with e(), since outreg is looking for e(), but mfx will save results in e() as well. Thus the need for holding and unholding the regression estimates. I also found the copy option on estiimates hold to be essential in this context.

Kit

*
* 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/