Dear users,
This question might sound simple or silly.
I have been trying to generate marginal effects after a logit regression with interactions. Becasue there are several interactions, I cannot use the -inteff- command.
I know that I can use the -predictnl- command but I have problems here in specifying the logistic function.
Below is an example of the case for a probit model with interaction:
probit foreign trunk mpg d /*note that d is a dummy*/
qui sum trunk
sca _tt=r(mean)
qui sum mpg
sca mg=r(mean)
local xb1 (_b[_cons]+ _b[trunk]*_tt + _b[mpg]*mg + _b[d]*1)
local xb0 (_b[_cons]+ _b[trunk]*_tt + _b[mpg]*mg + _b[d]*0)
predictnl dfdd=normal(`xb1')-normal(`xb0') in 1, se(serr)
list dfdd serr in 1
My question is: what density function do I specify for the logistic function if I use a logit instead of a probit?
Any suggestion will be welcomed.
Regards
Jon
*
* 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/
