Dear Statalist,
I have 2 questions regarding the margeff command (that estimates average
marginal effects) by Tamas Bartus. I am using Stata 8SE.
1. I get an error using the command but am not sure what this implies. I
checked the "help for margeff" (May 22 2005) and the Stata Journal
submission by Bartus (22 April 2005).
Here is what I do:
I run a probit on
continuous variables: pnursy2, age2, logatota2
dummies (0/1): ragender healthbad2
categorical variables: riadla2 radla2 hchild2 rlivsib2
dummies with several categories: race2 race3, edu2 edu3 and
square terms: agesq2 riadlasq2 radlasq2
Here is what I get:
probit rlivnhc6 pnursy2 age2 agesq2 ragender healthbad2 race2 race3
edu2 edu3 logatota2 riadla2 riadlasq2 radla2 radlasq2 hchild2 rlivsib2
if marp2==0&hchildf2==0, robust
Iteration 0: log pseudolikelihood = -1057.8623
Iteration 1: log pseudolikelihood = -927.72662
Iteration 2: log pseudolikelihood = -925.3291
Iteration 3: log pseudolikelihood = -925.32287
Iteration 4: log pseudolikelihood = -925.32287
Probit estimates Number of obs = 2054
Wald chi2(16) = 237.65
Prob > chi2 =
0.0000
Log pseudolikelihood = -925.32287 Pseudo R2 = 0.1253
------------------------------------------------------------------------
| Robust
rlivnhc6 | Coef. Std. Err. z P>|z| [95% Conf. interval]
----------+-------------------------------------------------------------
pnursy2 | .002027 .001292 1.57 0.117 -.0005052 .0045592
age2 | .0714462 .0074125 9.64 0.000 .0569179 .0859744
agesq2 | -.0011711 .0008612 -1.36 0.174 -.0028591 .0005169
ragender | .0972772 .089316 1.09 0.276 -.077779 .2723334
healthbad2| .1944082 .0768894 2.53 0.011 .0437077 .3451087
race2 | -.2096341 .1015799 -2.06 0.039 -.4087271 -.0105412
race3 | -.0055086 .226881 -0.02 0.981 -.4501873 .4391701
edu2 | .000135 .0849145 0.00 0.999 -1662944 .1665644
edu3 | .1068038 .0880652 1.21 0.225 -.0658009
.2794085
logatota2 | -.0208433 .0077476 -2.69 0.007 -.0360283
-.0056583
riadla2 | .6766688 .1953894 3.46 0.001 .2937126 1.059625
riadlasq2 | -.2465945 .0992594 -2.48 0.013 -.4411394
-.0520496
radla2 | .2762929 .1109397 2.49 0.013 .058855 .4937308
radlasq2 | -.083948 .0362853 -2.31 0.021 -.1550659 -.01283
hchild2 | -.0639309 .017976 -3.56 0.000 -.0991633
-.0286985
rlivsib2 | -.0305856 .0183409 -1.67 0.095 -.0665332 .005362
_cons | -6.227767 .600206 -10.38 0.000 -7.404149
-5.051385
---------------------------------------------------------------------------
. margeff, at(mean age2=78 agesq2=0 riadla2=0 riadlasq2=0 radla2=0
radlasq2=0) dummies(edu* race*) replace
Marginal effects at means on Prob(rlivnhc6==1) after probit
Variables set to specific values: age2 agesq2 riadla2 riadlasq2 radla2
radlasq2
estimates post: matrix has missing values
r(504);
I don't get this result if instead I use this command:
. margeff, at(mean age2=78 agesq2=0 riadla2=0 riadlasq2=0 radla2=0
radlasq2=0) dummies(edu2 edu3 race2 race3) replace
- but I am not sure if that gives me the correct results with respect to
the variables edu and race
2. Also, I was wondering if somebody knows how to determine the
relative weights on the variables in using the lincom command to get the
joint marginal effect for a variable which is included both by itself
and as a square term?
Thanks a lot for your help,
Kristin
Kristin J. Kleinjans
University of Aarhus
Department of Economics
Building 322
8000 Aarhus C
Denmark
Phone: +45 8942 1624. Fax: +45 8613 6334
E-mail: [email protected]
*
* 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/
