Hi Ged,
> Interesting point... this isn't actually the case normally, because SPM
> just uses a single global resels-per-voxel estimate in combination with
> the number of voxels in order to decide how many resels there are.
>
> You're right that small volume correction (SVC) does take the shape into
> account, but only if you specifically request SVC. Most people only do
> this if they are using small ROIs -- I guess arguably it should always
> be used, especially if the masking is quite severe (I've seen absolute
> thresholds of 0.2 advocated on this list for VBM, which, at least on my
> data, gives a very tight mask).
>
> What do other people think about this? Should tight/convoluted GM masks
> be treated with SVC like small ROIs are, or is there a good reason why not?
I remember a discussion along those lines a while ago (i.e., years :)
when I posed the same question ("wouldn't it be cool to only look at
fMRI results in gray matter") to Tom Nichols, and he somehow convinced
me that, no, it would not. Actually (I just found the mail, it was to me
and not to the list) he wrote:
> So, for the record, FDR makes no assumptions about the topology of
> the search region. Chop-chop-chop away. As long as the image used
> to determine your search region is orthogonal/independent from the
> experimential signal of interest (as the voxel grand means surely are),
> your fine.
>
>
> Let me know if the brain volume truncation helps. I had some
> students try it here and it didn't make that much of a difference.
So here you have an expert's opinion (as opposed to mine).
Best,
Marko
--
=====================================================================
Marko Wilke (Dr.med./M.D.)
[log in to unmask]
Universitäts-Kinderklinik University Children's Hospital
Abt. III (Neuropädiatrie) Dept. III (Pediatric neurology)
Hoppe-Seyler-Str. 1, D - 72076 Tübingen
Tel.: (+49) 07071 29-83416 Fax: (+49) 07071 29-5473
=====================================================================
|