Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: probit vs. logit
From
[email protected]
To
[email protected]
Subject
Re: st: probit vs. logit
Date
Tue, 25 May 2010 14:55:32 -0400
One reason that economists might prefer the probit conditional mean
specification is that it is easier to control for unobserved effects
via a Correlated Random Effects (CRE) approach, although this would
only arise in a panel data context. For example, Papke and Wooldridge
(1996) use the logit link function for cross-sectional analysis, but
in Papke and Wooldridge (2008) they use a probit link function to
allow them to more easily control for time-constant unobservables.
References:
1. Papke, L. and J. M. Wooldridge (1996). ?Econometric Methods for
Fractional Response Variables with an Application to 401(k) Plan
Participation Rates,? Journal of Applied Econometrics, 11, pp. 619-632.
2. Papke, L. and J. M. Wooldridge (2008). ?Panel Data Methods for
Fractional Response Variables with an Application to Test Pass Rates,?
Journal of Econometrics, 145, pp. 121-133.
Quoting Richard Williams <[email protected]>:
At 09:17 AM 5/25/2010, Nick Winter wrote:
I'll go further -- I've never seen a case where the choice of one
or
the other makes *any* substantive difference in the substantive
effects estimated. (That is, in predicted probabilities or in
the
impact of IVs on those probabilities.)
I'd also say that if you find an example where your conclusions do
differ, you would then be forced to make a choice based on theory
about the precise nature of unobserved disturbances -- I seriously
doubt most field have theories precise enough to make that choice
with any confidence.
I had a prominent economist tell me once that he thought probit was
much better, but he didn't say why. Like you, I've never seen a
case
where it made much substantive difference. It is more a matter of
disciplinary practices or personal preference. I don't know if it
is
true or not, but I think somebody on this list once said that you
would need several million cases to empirically distinguish between
logit and probit.
However, when you get into more advanced techniques, there can be
good reasons for preferring one link over the other. For example,
as
far as I know, there are no logistic counterparts to biprobit or
ivprobit. I don't know if that is because the routines have not
been
written or if such routines are impossible to estimate or much more
difficult to program.
-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME: (574)289-5227
EMAIL: [email protected]
WWW: http://www.nd.edu/~rwilliam
*
* 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/
*
* 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/