Dear Geraint,
Geraint wrote:
> Hi Jesper,
>
> > I don't really think it is meaningful in any way to look at main effect of age
> > using T2* weighted data. You could look at main effect of age using a
> > morphological technique (e.g. using T1 weighted and Voxel based morphometry)
> > or you could use a quantitative technique for measuring perfusion (i.e. PET or
> > perusion MRI).
>
> Just to clarify - we're talking about 'meaningful' in the sense of
> physiological interpretation rather than 'meaningful' in the statistical
> sense here? It doesn't seem to me that there is anything statistically wrong
> with comparing two sets of 'resting BOLD' data - just that it might be
> physiologically bogus.
>
> I'm thinking that it should be perfectly possible in (principle) to show a
> significant difference in resting T2* values between two study populations A
> & B. Of course, as you mentioned in your earlier mail, there are all sorts
> of confounds that would make the error variance for such a comparison high
> (scanner gain differences etc), but these would simply decrease the power of
> such a comparison rather than invalidate it. And if one showed such a
> significant difference, presumably it is likely to have *some* physiological
> correlate (we hope!)?
>
> If that's the case, then the issue is 'just' one of physiology i.e. what is
> the physiological correlate of 'resting BOLD'. Which would be open to
> empirical investigation. Or am I missing something here?!
>
You are absolutely right. It is a question of "meaningfulness" rather than
statistical correctness. If one is happy to discard the variance reduction achieved
from modelling session effects then it is certainly (statistically) kosher to
estimate main effect of age.
I would need to see a lot of "empirical evidence" though to believe that results
would in any way reflect neuronal activity. My bet is you will see mainly residual
anatomical differences and you would effectively be doing some really shoddy
morphometry on your T2* data.
>
> best,
>
> Geraint
>
All the best
Jesper
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|