Thread Subject: Negative Adjusted R-square

Hi,
Im getting negative adjusted R-square in my regression
analysis, though my R-square and other goodness fit
parameters look ok. I'd appreciate if you have any physical
meaning for this.

Thanks

On Jan 12, 6:17 am, "vicky " <vivek_muta...@yahoo.com> wrote:
> Hi,
> Im getting negative adjusted R-square in my regression
> analysis, though my R-square and other goodness fit
> parameters look ok. I'd appreciate if you have any physical
> meaning for this.
>
> Thanks

R^2>0 if there is a constant term in your regressors


Regards,
Sanya

You probably have too low of a ratio of observations to regressors. The expected R square by chance is k/(n-1), so for example if you have 12 regressors and a sample size of 30 expected R square by chance is .414. When R squared is less than the expected R square by mere chance, adjusted R square can be negative. You most likely have some useless regressors in you model.

"Dr. Stat" <marvin.wolverton@sbcglobal.net> wrote in message
<6286268.1200269479230.JavaMail.jakarta@nitrogen.mathforum.org>...
> You probably have too low of a ratio of observations to
regressors. The expected R square by chance is k/(n-1), so
for example if you have 12 regressors and a sample size of
30 expected R square by chance is .414. When R squared is
less than the expected R square by mere chance, adjusted R
square can be negative. You most likely have some useless
regressors in you model.

Thanks for your suggestions. You are right. Im using about
92 regressors for 95 sample size, and my Rsquare < expected
R-square. i presume there are large useless regressors
making adjusted R-square negative.