Hi
There is no conclusion as such, interpretation is different between the two in precisely the same way as it's different reporting max Z or max %BOLD in task FMRI. What you want to do depends on the question you want to ask and is not context invariant.
hth
Christian
On 7 Mar 2011, at 13:06, AnaBea Solana wrote:
> Good morning
>
> In this thread, it was debated the possible different benefits or merits of using the raw (effectively PE) images dual-reg output (after TC-GICA) by stage 2 into randomise, or the zstats.
>
> We were wondering if you have already any conclusion about this topic.
>
> thank you
>
> AnaBea
> On Tue, Oct 19, 2010 at 5:13 PM, David V. Smith <[log in to unmask]> wrote:
> Thanks, Steve!
>
> >> (or should this be done using the *_Z.nii.gz images?
> >
> > This is a good question - whether to feed the raw (effectively PE) images output by stage 2 into randomise, or the zstats - either is valid, and we're currently looking into the relative merits of each.
> >
>
>
> Good to know both approaches are equally valid, and I look forward to hearing what you guys think about the relative merits of each. Using the z-stats instead would seem to have the benefit having a normalized value for each subject, or at least having values that have been scaled by their uncertainty/noise, which might reduce the impact of outliers. But I don't know enough about this to have an informed opinion here, so I'll leave it to the experts. Until then, I'll use the raw images since it looks like this what you guys have been using by default in the DR script.
>
>
> Cheers,
> David
>
|