I am still having an issue with svy:probit that I cannot resolve. I've
attempted to replicate it here using the nhanes data. The issue is the
discreptancy in the significance of the marginal effects (both mfx and
margeff) and the coefficients from the probit. My intuition is that
they should be the same. In my own data, I get lots of significant
coefficients that don't have significant average marginal effects.
While I was unable to replicate this with the nhanes data, here's an
example (the code is below). The blood pressure dummy is not
significant with the beta criterion, but it is with the marginal
effects. I am really not sure how to interpret this.
betas mfx margeff
height (cm) -0.0078 -0.0002 -0.0006
(0.0065) (0.0002) (0.0005)
weight (kg) 0.0066** 0.0002** 0.0006**
(0.0028) (0.0001) (0.0002)
1=female, 0=male -0.4447*** -0.0127*** -0.0380***
(0.1051) (0.0036) (0.0091)
1 if race=black, 0 otherwise -0.0261 -0.0007 -0.0022
(0.0863) (0.0022) (0.0070)
age in years 0.0425*** 0.0012*** 0.0036***
(0.0027) (0.0001) (0.0003)
1 if BP > 140/90, 0 otherwise -0.1418 -0.0034* -0.0110*
(0.0842) (0.0018) (0.0060)
Constant -3.0356**
(1.1088)
Observations 10335 10335 10335
Standard errors in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
Here is the code:
use "http://www.stata-press.com/data/r9/nhanes2f.dta";
svy: probit heartatk height weight female black age highbp
outreg using file.out, ctitle("Betas") bdec(4) coefastr se 3aster replace
local obs = e(N)
mfx
mat margeff = e(Xmfx_dydx)
mat standerr = diag(e(Xmfx_se_dydx))
mat variance = standerr*standerr
ereturn post margeff variance, obs(`obs')
outreg using file.out, ctitle("mfx") bdec(4) coefastr se 3aster append
svy: probit heartatk height weight female black age highbp
margeff, at(mean) model(probit) replace dummies(female \ black \ highbp)
outreg using file.out, ctitle("margeff") bdec(4) coefastr se 3aster append
*
* 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/
