dated Tue, 14 June 2005 9:48:54-- 0700
message from
Hyojoung Kim" <
[email protected]>
Hyojoung Kim asks in an earlier message if there is a formal test for
determining which (logistic or probit) is more appropriate than another.
Wouldn't comparing one or more of the the AIC, BIC, and log-likelihood
values after doing generalized linear modeling help at this ? The two
commands would then presumably be:
glm y x, fam(bin) link(logit)
versus
glm y x, fam(bin) link(probit)
I had thought that this was a more or less standard methodology for
comparing models and assessing which model specification might be
preferred. Am I right in this. Presumably, it would also be
advisable to regress predictions from the logit model against
predictions from the probit model and do a scatter plot to evaluate how
closely they agree.
david.
David J. Miller, Chief
Chemistry & Exposure Branch
Health Effects Division
Office of Pesticide Programs
703-305-5352 (voice)
703 605-1289 (fax)
visit: http://www.epa.gov/pesticides/
*
* 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/
