Dear statalist,
I tried to replicate marginal effects after xttobit, re by hand. When
estimating tobit, this is no problem:
tobit y x1 x2, ll(0)
su x1 if e(sample)
local mean_x1=r(x1)
su x2 if e(sample)
local mean_x2=r(x1)
local xb "_b[x1]*`mean_x1'+_b[x2]*`mean_x2'+_b[_cons]"
nlcom normden((`xb')/_b[/sigma])*(_b[x1]/_b[/sigma])
produces the same command as mfx, predict(pr(0,.)).
When estimating xttobit y x1 x2, i(id) ll(0)
and substituting sigma by sigma_e (thus assuming sigma_u to be zero),
however, this produces different results as compared to the mfx command. Any
ideas what I do incorrectly?
Any help is much appreciated!
Best,
Axel Dreher
*
* 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/
