<>
Dear Statalisters,
a colleague of mine recently complained about a supposed bug in Stata. He
ran a -tobit- estimation and wanted to obtain the marginal effects
post-estimation. Curiously, plain -mfx- w/o options returned the parameter
estimate from the -tobit- estimation, as in
********
sysuse auto, clear
generate wgt=weight/1000
tobit mpg wgt, ll(17) ul(24)
mfx
********
It seemed to him as if -mfx- added no information. Now, I know this is
correct as -mfx- is designed to use the default method from -predict- if the
user does not specify what he/she wants. But it was not as obvious to him.
-h tobit postestimation- does provide the correct commands
********
mfx compute, predict(e(17,24))
mfx compute, predict(ystar(17,24))
********
My question is thus: Should the dialog box for -mfx- change according to the
preceding estimation command, just as -predict- does? That would facilitate
the task for many casual users, but I am aware that it would also tie up
resources at StataCorp to program the additional boxes. And should the
default after -tobit- really be the linear prediction which Baum (2006, p.
264) calls "rarely useful"?
Best
Martin
Baum, Christopher, An Introduction to Modern Econometrics Using Stata, Stata
Press, 2006
*
* 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/
