Dear Tetsu,
> Dear SPM
>
> Quick and basic question regarding this; (see below)
>
> Are both of ANCOVA and proportional scalling using mean adjusted
> value (signal intensity adjusted to mean of 100 for fMRI and
> radioactive count adjusted to mean of 50 for PET) for subsequent
> statistical analysis?
>
> Thanks
>
Both ANCOVA and proportional scaling attempt to reduce the variance
attributed to global changes, thereby potentially increasing the
sensitivity. Strictly speaking only proportional scaling adjusts data
prior to "subsequent statistical analysis", whereas ANCOVA includes the
globals as a covariate in the analysis. Although proportional scaling
and ANCOVA are based on different assumptions most empirical
comparisons suggest that quite similar results are obtained with the
two methods.
Good luck Jesper
> >At 11:06 PM 9/10/98 -0400, you wrote:
> >| Interestingly, using proportional scaling for normalization (but without
> >| any scaling of the grand mean) gives me exactly the same values. So
> Global
> >| Normalization="None" and Global Normalization="Proportional Scaling" look
> >| like they are doing the same thing. How can that be?
> >
> >Proportional scaling scales each image to have the specified global mean.
> >Since this isn't a particularly interesting global value to look at, SPM
> >prints out the unscaled globals, as it does when no proportional scaling or
> >grand mean scaling is selected.
> >
> >-andrew
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|