Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: xtabond and rsquare goodness of fit measure


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: xtabond and rsquare goodness of fit measure
Date   Fri, 15 Aug 2003 14:48:15 +0100

Parthiban David
>
> I estimated a regression model using xtabond in stata. A
> referee wants us to
> report a goodness of fit measure (e.g., rsquare or pseudo
> rsquare). I
> checked the stata archives (using findit rsquare) but could not find
> anything specific for xtabond. The closest was an faq by
> William Gould which
> presents a procedure to compute pseudo Rsquare for probit.
>
> Does anyone know of an equivalent way to compute a pseudo
> Rsquare for
> xtabond?
>
> I noticed that xtabond reports chi-squared statistics for
> the variables
> present in the model. Does it make sense to compute a
> goodness of fit by
> running two separate xtabond regressions using a full model
> and reduced
> model and comparing the chi-squared statistics?

This may be stupid in view of specifics of -xtabond-, but
when all else fails one R^2 measure is always obtainable as
the square of the correlation between the observed and
the predicted, the latter being always calculated
in the same form as the response. In other words,
use -predict- to get predicted, and then -correlate- and
finally square.

Make sure you restrict calculation to the estimation sample.

This is how I would do it for -regress-

regress <response> <whatever>
predict predict if e(sample)
correlate <response> predict
di %23.18f (r(rho))^2

Nick
n.j.cox@durham.ac.uk

*
*   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