I'm not sure if r squared is a meaningful measure with robust
regression. See for example the following posts, which suggest to me
that it is not:
http://www.stata.com/statalist/archive/2004-05/msg00107.html
https://stat.ethz.ch/pipermail/r-help/2008-November/179773.html
cheers,
-MH
On Thu, 2011-08-04 at 22:17 +0100, Michael Lombardo wrote:
> Hi FSLers,
>
> I have a stats question that isn't necessarily related to things
> implemented by FSL, but is about applying matlab code on some imaging
> data I have. I'd like to calculate a partial r squared value or
> coefficient of determination using robust regression (my current
> analysis uses the Matlab function robustfit to do the robust
> regression). Could anyone help with a suggestion about how to do
> this? I know how to do this in an ordinary least squares regression,
> but am at a loss right now for what is the proper way to do this for
> robust regression.
>
> Thanks,
>
> Mike
|