Hi,
> Some authors think the methods for correction of multiple comparisons implemented in spm99 are too conservative for VBM studies in contrast with functional imaging.
> I wonder where is the problem.
Well, there are two problems. One is that the random field correction
for small degrees of freedom (as often in VBM) _is_ too conservative;
you can correct that by using non-parametric statistics (SnPM), or, if
you are brave, FDR.
The second is that low degrees of freedom mean low power, but there's
nothing much to be done about that, if you want a valid statistical
threshold for control of false positives.
> Are there any changes in correction using random field theory between spm99 and spm2?
> Did someone make computations of this problem to quote it?
There are some small changes, but the big differences are likely to
come from FDR vs random field correction.
> The second question I want to ask is about small volume correction. Mathew Brett wrote the formula for computing ECs in small volumes have to be adjusted for shape differences. So, you get different results when you do a ROI analysis (for example hippocampus mask using WFU PickAtlas) and when you do small volume correction with the same mask as in ROI.
> What is more appropriate aproach?
> Is there something that favours one of the two aproaches?
I just added a new entry to the MarsBaR FAQ:
http://marsbar.sourceforge.net/faq.html#svc
- I hope it is useful...
Best,
Matthew
|