Dear Stuart,
Just a few comments (from a non-statistician).
>My understanding is that a bonferroni
>correction within SPM is overly harsh,
It's perhaps worth mentioning again that SPM doesn't use a Bonferroni
correction (see Tom Nichols' post on Sunday).
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0101&L=spm&F=&S=&P=20339
>Firstly, by the principle of materialism always being correct, it has
>to be the case that our experimental interventions alter activity in
>the brain.
Sure, but they don't necessarily have to alter the BOLD signal
significantly (if that's what we're measuring).
[Although, admittedly, if the BOLD signal is a continuous variable,
then presumably the a priori probability that its underlying value is
exactly zero is itself zero. So I suppose that in theory the null
hypothesis of 'activation = zero' is inevitably false. But...]
>The null hypothesis is always wrong, the profile of
>activity has to change.
It doesn't matter if the null hypothesis is wrong! The job of the
functional imager is to PROVE that it is wrong (usually in a
particular direction), and unless you provide sufficient evidence to
reject the null hypothesis, you haven't done this.
>If you are searching for a regional effect
>then the story changes (although there is plenty of BS to be had
>between "changes in the brain" and "changes in regions x,y,z"). If
>you are looking for a particular region or network of regions then it
>would be advisable to calculate error rates so as to assess the
>possibility of a type II error.
I guess that you would only correct over a 'small volume' in this instance.
Best wishes,
Richard.
--
from: Dr Richard Perry,
Clinical Lecturer, Wellcome Department of Cognitive Neurology,
Institute of Neurology, Darwin Building, University College London,
Gower Street, London WC1E 6BT.
Tel: 0207 679 2187; e mail: [log in to unmask]
|