<>
John wanted to do nonlinear predictions from an estimated logit model.
Wikipedia (or Wolfram's MathSource) is his friend:
sysuse auto,clear
logit foreign trunk mpg d
// mean of predictions is equal to mean of foreign in logit
su foreign, mean
scalar mu = r(mean)
su trunk, mean
scalar mut = r(mean)
su mpg, mean
scalar mum = r(mean)
// form of CDF from http://en.wikipedia.org/wiki/Logistic_distribution
scalar s = sqrt(2*c(pi)/3)
local cum1 1/(1+exp(-((_b[_cons]+ _b[trunk]*mut + _b[mpg]*mum +
_b[d]*1) - mu)/s))
local cum0 1/(1+exp(-((_b[_cons]+ _b[trunk]*mut + _b[mpg]*mum +
_b[d]*0) - mu)/s))
di "`cum1'"
di "`cum0'"
predictnl dfdd1= `cum1' - `cum0' in 1, se(serr1)
l dfdd1 serr1 in 1
Kit Baum | Boston College Economics & DIW Berlin | http://ideas.repec.org/e/pba1.html
An Introduction to Stata Programming
| http://www.stata-press.com/books/isp.html
An Introduction to Modern Econometrics Using Stata | http://www.stata-press.com/books/imeus.html
*
* 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/
