Gang Wrote:
-------------------------------------------------------
> Hi Michael,
>
>
> > If we take the square root of cov1_r+tlrc, is
> the partial correlation coefficient for covariate
> 1
> > while controlling for covariate 2?
>
> Do the two files give you the same or different
> results? This is what it says in Doug Ward's
> manual for 3dRegAna: "R^2 is an indicator of the
> efficacy of the overall regression, and does not
> indicate the significance of the individual
> parameter".

Yes, the two files are equal. So, R^2 is for the overall regression model and not partial. The value is equal if I use the -brick m rstat option, which is the coefficient of multiple determination R^2

>
> > if so, if we also generated the -tcoef k file,
> will the sign of the -tcoef coefficient/t
> determine
> > the sign of the partial correlation coefficient?
>
>
> According to the manual, the R^2 is not really the
> partial determination coefficient. However, you
> can directly obtain the partial correlation
> coefficient based on the t-value using the
> conversion formula on the following page:
>
> [afni.nimh.nih.gov]
>

Thanks! This is exactly what we want.


> > Is there an easier way to generate these files
> into one file? 3dRegAna outputs these files
> separately.
>
> You can use option -bucket to bundle everything
> into one file.
>
> > Is it possible to generate the same results
> using 3dttest++ or another program?
>
> Yes, you can use 3dttest++ to get everything
> except for R^2.

Thanks again.