Hi Rachel,

I believe both options (case studies vs a group RFX analysis) are valid. In
this case, I would rather go with the case studies, because the brain damage
in your patient group increases to a large extent the variability that is
already there in subjects without brain damage. Depending on lesion site and
size, there may be quite different functional strategies these brains employ
to maintain performance, subserved by different neural networks. These are
currently under investigation in many labs. Another thing to take into
account is the normalization required for a group study, which is not
straightforward for lesioned brains - you can do a very good job, with
masking the lesion and/or normalizing to gray matter etc., but especially
perilesional activity will always be difficult to compare over differently
lesioned individuals, if you pool them together.

So, why not do both? You could present a study comparing the succes of TMS
in regions identified in case studies, vs regions identified through a RFX
group analysis.

I, for one, would be very interested in the results!

Best wishes, Dirk
************************************
Dirk-Bart den Ouden, Northwestern University

Aphasia & Neurolinguistics Research Laboratory
Dept. of Communication Sciences & Disorders
2240 Campus Drive, Evanston IL 60208-3066
Phone (lab): 1-847-467-7591

************************************

-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On
Behalf Of Rachel Mitchell
Sent: Monday, June 05, 2006 11:18 AM
To: [log in to unmask]
Subject: [SPM] Fundamental problem with group averaging?

Hello list subscribers.

I wish put to you what seems to be quite a fundamental issue.
I would be interested for ANY thoughts and comments on this matter.

I am currently planning a combined TMS and fMRI study.
The scientific issue is that differences are emerging, between the brain
regions that mediate function X as suggested by the older lesion studies,
compared to those brain regions suggested by the newer fMRI studies.
One way to resolve this issue seems to be to incorporate TMS, to selectively
target the areas suggested by the fMRI data.

So, there I was, planning what fMRI paradigms I was going to run.
I would then identify the active brain regions suggested by the group
average and  in phase 2, use TMS to target these areas, to observe their
behavioural effects.

Then along came someone I often have email discussions with, who said I
should NOT do this.
What I needed to do he said was to do this on an individual basis...scan
person 1, look at their individual fMRI response, then target TMS on that
individual in those person specific brain areas.
Adopting this kind of approach instead could cause quite profound practical
issues.
He pointed me towards papers such as those by Steven Small which showed that
on a language task the localization varied amongst individuals. When he made
a "composite localization" for the group, it was not in the region activated
by any of the individual subjects.

My concern is what this implies about group averaging.
If a novice considered the above snippet, you might be forgiven for
beginning to think group averaging was flawed. 
However, random effects modelling as in SPM considers both within- and
between-subject variance, so why does this situation occur?
The purpose of the random effects analysis is supposed to be to find the
areas that are activated in much the same way in all subjects, as opposed to
fixed effects models which give you areas that are activated "on the
average" across the subjects.

My experience is limited to SPM, so I do not have the technical expertise to
dissect the analysis procedures used in the study cited above (which only
seemed to use SPM for the group analysis), but if one has used the classical
two stage SPM random effects analyses, shouldn't you be protected to some
extent from this kind of thing happening.

I'd be interested to know the thoughts of people who know about random
effects, and also those of others who have combined TMS and fMRI in their
research.

With very best wishes
Rachel

--------------------------------------------------------
Dr Rachel L. C. Mitchell.
Lecturer in Cognitive Psychology, University of Reading.
Honorary Research Fellow, Institute of Psychiatry.
Research Psychologist, Berkshire Healthcare NHS Trust.

Correspondence Address:
School of Psychology and Clinical Language Sciences, Whiteknights Road
University of Reading Reading Berkshire
RG6 6AL

Tel: +44 (0)118 378 8523
Direct Dial: +44 (0)118 378 7530
Fax: +44 (0)118 378 6715
-----------------------------------------------------------------------