Maximum number of iterations exceeded with -predictnl-

[Xiaojin Wang]

Dear all,

I am using -predictnl- to calculate marginal effect of a dummy variable for a natural logged dependent variable after -Heckman-.

predictnl ME_size3=exp(_b[#1:3.household_size_R] + `beta_lambda' * ///
(((normalden(xb(#2)-`household_size_R') + _b[#2:3.household_size_R]*1))/normal(xb(#2)-`household_size_R' + _b[#2:3.household_size_R]*1) ///
- (normalden(xb(#2) -`household_size_R')/normal(xb(#2)-`household_size_R'))))-1, se(Se_ME_size3) p(sig_ME_size3)

*Get ERROR message: Maximum number of iterations exceeded.

I then specify the iteration number to be 1000, get the same message.

I read from an old post about the same problem: http://www.stata.com/statalist/archi.../msg00585.html, in which rescaling the variables is recommended.

I adjust my code to be:

predictnl ME_size3log=(_b[#1:3.household_size_R] + `beta_lambda' * ///
(((normalden(xb(#2)-`household_size_R') + _b[#2:3.household_size_R]*1))/normal(xb(#2)-`household_size_R' + _b[#2:3.household_size_R]*1) ///
- (normalden(xb(#2) -`household_size_R')/normal(xb(#2)-`household_size_R')))), se(Se_ME_size3log) p(sig_ME_size3log)

It worked, I can get the desired marginal effects

predictnl ME_size3 = exp(ME_size3log) - 1, se(Se_ME_size3) p(sig_ME_size3)

I wonder by doing the calculation in two steps, do I get the correct standard errors, and significance levels?

Thanks.