Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Pseudo R2 - Tobit regression

9 views
Skip to first unread message

mac

unread,
Dec 22, 2009, 6:36:44 PM12/22/09
to
Hello,

In the output of the tobit model it only appears the Log likelihood,
however
we will like to compare two models, therefore we will like to have a
measure
of the goodness of fit like the pseudo R2. Is there any way to obtain
it
automatically or shall we run the model just with a constant? How can
I calculate the McKelvey-Zavoina R2? For this porpuse(McKelvey-Zavoina
R2) I have tried to calculate the statistic but I must be doing
something wrong because the result set are very different from those
shown by OLS-R2.

Thank you

mac

unread,
Dec 23, 2009, 8:32:00 AM12/23/09
to

Sorry, I forget that I work with LIMDEP

Rich Ulrich

unread,
Dec 23, 2009, 6:38:59 PM12/23/09
to
On Tue, 22 Dec 2009 15:36:44 -0800 (PST), mac <macar...@gmail.com>
wrote:

You want AIC (Akaike information criterion) or BIC, which
are devised for the purpose of comparisons. However much
you trust them.

Estimates of "Pseudo R^2" are attempts to find a scale
similar to OLS. That leaves them a step away from being
something already adapted for comparisons with each other.

--
Rich Ulrich

0 new messages