Re: st: The speed of the MFX command and Stata 9

[Thomas Corneli�en <cornelissen@mbox.iqw.uni-hannover.de>]

Hi Larry,
I have made the experience that -mfx- has been considerably speeded up in 
Stata 9.
My experience relies on -mfx- after the heteroscedastic probit 
model -hetprob-. For example, in one model Stata 8 took 3:24 hours for -mfx 
compute- after -hetprob-, while Stata 9 took 1 minute 24 seconds. A larger 
model hadn't finished after 39 hours in Stata 8, but it took only 10 minutes 
in Stata 9.

However, I have no experience with -mfx- after -intreg-, so I couldn't tell 
you for sure it would also be quicker. Maybe other users have experience 
with this.
Thomas

-------------------------------------------------------------------------
Thomas Cornelissen
Empirical Economic Research
University of Hannover, Germany



Larry worte:
-----------------------------
Hi,

I have been attempting to run a fixed effects tobit model (despite some
of the pitfalls involved).  The model involves 11,000 villages in 850
counties.  I regressed the log of my Y variable on a set of about 25
continuous and dummy variables and also 850 dummy variables representing
the counties.  I use -intreg- for the regression in Stata SE 8.2 for
UNIX (with a graphical interface through Reflection X).  This takes
about 1.5 hours to complete.  After this estimation I typed - mfx
compute, predict(e(0,.))- since my Y variable is censored at 0.  This
command ran for more than 5 days with no results before my computer
crashed (not related to the Stata process).

My question is would Stata 9 SE (on a 3.4 Ghz Windows machine) run this
appreciably faster than Stata 8 SE or am I just asking too much of
Stata?  I am a grad student so I am trying to wait to upgrade to Stata 9
until I have a job and hopefully a research budget, but since I need to
finish my job market paper in the next 2 months I would gladly upgrade
now if there is a chance Stata 9 would solve my problem.

I did try the command -margin, mean dummies(kec_code1-kec_code841)- but
I got pages and pages of error codes involving the fixed effects that
looked like "s(kec_code841) : "if kec_code1!=1 & kec_code2!=1 &
kec_code3!=1 & kec_code4!=1 & kec_code5!=1 & kec_code6!=1........"

Thanks for any help you can give.
Larry Chavis
PhD Candidate
Stanford Graduate School of Business 
*
*   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/