Hi Marko,
Thanks for the advice - this was more or less the conclusion I had come
to. I was however confused because looking at the mailing list, the
discussion regarding SVC in August suggested that the threshold chosen
initially was important when carrying out a SVC. Maybe I just
misunderstood this discussion !
Mahinda
Marko Wilke wrote:
> Hi Mahinda,
>
>> Could someone clarify the following for me ?
>
> Not sure, it is actually more a question I have :)
>
>> Following from the recent SVC discussion I was under the impression
>> that the SVC when applied only includes those voxels (within whatever
>> volume you have defined) that were initially activated at the
>> previously chosen uncorrected threshold. In which case the value
>> chosen for this threshold should dictate the number of voxels within
>> the region where the SVC is applied and therefore also the resulting
>> p value for the analysis. However, I find that regardless of the
>> initial threshold chosen (ie be it 0.01 uncorrected or 0.99
>> uncorrected) the resulting p value when the SVC is applied (and the
>> same region is used) is the same. This would suggest that a small
>> volume correction includes all those voxels within the pre-defined
>> area that one is analysing regardless of what threshold was
>> originally chosen for the uncorrected analysis. Is this is correct,
>> because reading the emails from August on this subject, I did not
>> think this was the case.
>
> Without remembering what the details of this exchange were, I would
> intuitively expect the behavior you described.
>
> The rationale for a small volume correction, to me, always was that "I
> expect activation within a certain region", as defined by the search
> volume. Therefore, the correction for multiple comparisons should take
> into account the volume of this region, and not of activation seen
> within this region. I may be wrong but taking only the activated
> voxels of a prior analysis would be post-hoc, which really is not what
> "having an a priori hypothesis about the region of activation" would
> be about. Right ?
>
> Also, if you define a search region for a small volume correction, it
> really does not matter if there were activated voxels in the first
> place as they may only become significant after SVC, which is
> perfectly fine (and likely what most people are using it for :) I
> would see no good rationale for excluding voxels within the search
> volume on the basis of an analysis that you are not interested in
> anyway (as it corrected for the whole brain), but there may be factors
> involved that I do not see (anyone: are there?). This is just my take
> on things which may be perfectly wrong, but perhaps it stimulates
> somone who actually knows it to correct me.
>
> Best,
> Marko
|