Hi Pierre, hi all,
Sorry, that I have to disagree with this last statement by Pierre:
>- finally, the observation of each subject separately is a key
>verification. If the effect is observed for each subject (even at a
lower
>threshold and a location slightly different of the group location), I
think
>that the statistical test (either FWE, FDR or any other test) on the
group
>make sense even if the p value is relatively high (p=0.05 and a
moderate
>effect in 10/10 subjects seems to me more credible that a p=0.001 and
the
>effect observed in 5/10 subject).
If no initial threshold is required in determining whether there is an
effect
in an ROI (ROIs consist typically of 400 or more voxels) and the
location is
arbitrary (e.g. any of the voxels/resels within the ROI) then selecting
the 'effect' for each subject means to select the maximum activation
value amongst a set of values.
I have seen articles where then a statistic (mean) is computed over
these
subject-separate maximum values. This is not valid in my view,
for the following reason:
If the set of activation values within the ROI were randomly
distributed
with a mean of zero you will easily find extreme activation values
in this set (the expected number depending on the variance of the
distribution).
So this creates an effect out of noise.
I would not say that this method were totally invalid, but the
statistic has to
take into account the expected number of maximum values under the
null hypothesis.
I am not a statistician, so I would like to get the views of others on
this.
Best
Anja
Dr. Anja Ischebeck
Innsbruck Medical University
Clinical Department of Neurology
Anichstrasse 35
A-6020 Innsbruck - Austria
tel.: +43 (0) 512 504 23661
|