Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: probit questions


From   "Verkuilen, Jay" <JVerkuilen@gc.cuny.edu>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: probit questions
Date   Wed, 25 Jun 2008 13:06:51 -0400

Maarten buis wrote:

>>The differences are very minor, I believe that the tails are slightly
fatter for the logistic distribution than the normal (or was it the
other way round?),<<

Nope, you're right. The logit is like a t(10). The Cauchy is, of course,
a t(1). Thus, in terms of kurtosis, probit < logit < cauchit. (I don't
think Stata has a built-in cauchit link but -gologit2- does, if I recall
correctly.) Roger Koenker had a very nice paper on this:
http://www.econ.uiuc.edu/~roger/research/links/links.html 


>>anyhow you won't find any difference with only 60 observations. 

Nope. 


>>So your choice should be quided by convenience and conventions in your
discipline. I like -logit- because I like odds ratios, but economists
tend to like -probit-.<<

Economists seem like probit because they like the normal distribution
for random utility models. Of course, you can use a logit random utility
model, too, if the random utilities are Type I extreme value. The
multivariate normal is "nicer" in some respects because of its closure
properties. 

JV

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index