As far as I know, in nonlinear regression it is possible when calculating R2
from 1-(SSE/SST) that the residual SS (SSE) is greater than SST... Thus
creating a negative R2 value. I have read that these types of R2 values
should just be set to 0. Is this the case? I would like to have some sort
of explanation as to whether or not this is an acceptable approach... And
why or why not?
Thanks.
Kevin.
-----Original Message-----
From: Graham Wright [mailto:
[hidden email]]
Sent: October 20, 2006 11:56
To: Kevin Bladon
Cc:
[hidden email]
Subject: Re: Negative R2 - Nonlinear regression
Are you talking about the Adjusted R2? The regular R2 should never be
negative (unless r is a complex number!), but if your adjusted R2 is
negative that means that your model has a lot of variables that don't
contribute any explanatory power to the model, and/or that the
explanatory power of the model in general is negligible.
-Graham
Kevin Bladon wrote:
> Calling all statisticians...
>
> Can someone please explain, what would be the most appropriate way to deal
> with a negative R2 in nonlinear regression?
>
> Set it to 0? What would be the resoning for this approach?
>
> Regards,
> kevin.
>
Locked
