# st: margeff and margeff at the mean differences

 From Nikos Nikiforakis To statalist@hsphsun2.harvard.edu Subject st: margeff and margeff at the mean differences Date Wed, 14 Feb 2007 12:21:14 +1100

Dear members of Statalist,

I am estimating the average marginal effects following a probit with random effects using -margeff-. My 'problem' is that the average marginal effect of one variable (x2) turns out to be not significant, although it is significant at the 1-percent level in the probit and also when I use -margeff- to estimate the marginal effects at the mean. (The marginal effects of all other variables are very similar).

Any idea why this might be happening. I would be grateful for any suggestions as I could not find an answer in Bartus (2005, SJ) which introduces -margeff- or elsewhere.

Thank you in advance
Nikos

Here is the Stata output (x1 x3 and x4 are dummy variables)

. xtprobit y x1 x2 x3 x4 x5, i(cluster)

Fitting comparison model:

Iteration 0: log likelihood = -2759.006
Iteration 1: log likelihood = -2267.4507
Iteration 2: log likelihood = -2257.582
Iteration 3: log likelihood = -2257.5472

Fitting full model:

rho = 0.0 log likelihood = -2257.5472
rho = 0.1 log likelihood = -2196.2157
rho = 0.2 log likelihood = -2201.078

Iteration 0: log likelihood = -2195.3866
Iteration 1: log likelihood = -2191.6554
Iteration 2: log likelihood = -2191.6139
Iteration 3: log likelihood = -2191.6139

Random-effects probit regression Number of obs = 5760
Group variable (i): cluster Number of groups = 32

Random effects u_i ~ Gaussian Obs per group: min = 120
avg = 180.0
max = 360

Wald chi2(5) = 775.44
Log likelihood = -2191.6139 Prob > chi2 = 0.0000

------------------------------------------------------------------------------
y | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
x1 | -.7448295 .1461338 -5.10 0.000 -1.031246 -.4584126
x2 | .1511212 .0059109 25.57 0.000 .139536 .1627064
x3 | -.0558861 .1622825 -0.34 0.731 -.373954 .2621818
x4 | -.0024064 .1447156 -0.02 0.987 -.2860437 .281231
x5 | -.0769059 .0076225 -10.09 0.000 -.0918457 -.0619662
_cons | -.5891004 .1766414 -3.34 0.001 -.9353111 -.2428897
-------------+----------------------------------------------------------------
/lnsig2u | -1.93502 .322295 -2.566707 -1.303333
-------------+----------------------------------------------------------------
sigma_u | .3800282 .0612406 .2771065 .5211765
rho | .126196 .0355397 .0713121 .2136046
------------------------------------------------------------------------------
Likelihood-ratio test of rho=0: chibar2(01) = 131.87 Prob >= chibar2 = 0.000

. margeff

Average partial effects after xtprobit
y = Pr(y)

------------------------------------------------------------------------------
variable | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
x1 | -.1497686 .022374 -6.69 0.000 -.1936208 -.1059164
x2 | .0007352 .001196 0.61 0.539 -.0016089 .0030793
x3 | -.0113159 .0324428 -0.35 0.727 -.0749025 .0522707
x4 | -.0004869 .0292553 -0.02 0.987 -.0578262 .0568523
x5 | -.0155627 .0017951 -8.67 0.000 -.019081 -.0120444
------------------------------------------------------------------------------

. margeff, at(mean)

Partial effects at fixed values after xtprobit
y = Pr(y)

------------------------------------------------------------------------------
variable | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
x1 | -.1617599 .0343461 -4.71 0.000 -.2290771 -.0944427
x2 | .0326642 .0031176 10.48 0.000 .0265538 .0387747
x3 | -.0120799 .0355729 -0.34 0.734 -.0818015 .0576417
x4 | -.0005201 .0312798 -0.02 0.987 -.0618274 .0607872
x5 | -.0166266 .0021715 -7.66 0.000 -.0208826 -.0123706
------------------------------------------------------------------------------

Nikos Nikiforakis
Research Fellow
Dept. Economics
The University of Melbourne
http://www.economics.unimelb.edu.au/staffprofile/nnikiforakis.htm

Nikos Nikiforakis
Dept. Economics
The University of Melbourne
http://www.economics.unimelb.edu.au/staffprofile/nnikiforakis.htm
*
* 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/