In 1999, Paul Allison published a paper in Sociological Methods &
Research (vol. 28 issue 2 pages 186-208) about comparing logit
coefficients across models. I am trying to use the Stata code that he
included in the paper's appendix (see below). In particular, I want to
compare coefficients for two logistic regression models, one with men
and one with women, but I can't get the code to run properly.
Stata 8 seems to understand and define the program below, but the ml
maximize statement does not run, and it produces the following error:
"*__00000S invalid name ". I've checked and double-checked the program
and my modified version of the model statement in the belief that I have
simply typed something wrong, but I can't seem to figure out where the
problem is.
Has anyone used Allison's program successfully? Any idea why the code
below does not run?
Maybe my problem is with the maximize statement and not the program?
Thanks,
Jeremy
--Below I have listed the Stata code and comments from the appendix--
program define glogit
version 6
args lnf theta delta
quietly replace
`lnf'=$ml_y1*`theta'*(1+`delta')-ln(1+exp(`theta'*(1+`delta')))
end
ml model lf glogit (prom = undgrd arts dur dur2 prest female) (delta:
female,nocons)
ml maximize
The commands bracketed by PROGRAM and END define the estimation problem
in a general way. These commands can be saved in a DO file and used for
any data set or particular set of variables in the model. In the ML
MODEL command, LF specifies the optimization algorithm and GLOGIT refers
to the name of the program defined earlier.
