Dear Dear Dr. Holmes, Dr. Zahrahn, and other SPMers,
I will attempt to simplify my last e-mail which I think may help me get a
reply. I would be grateful for any response. (I apologize for my paranoia
but I have never received a response from any of my past spm-mailbase
mailings).
I have a question regarding making a group z-map from 7 individual
subject z-maps task minus rest condition z-maps.
I want to know if the following approach is reasonable and/or has
any glitches.
At each Talairach coordinate, I perform a t-test on the
z-scores from each subject generating a probability (and t-score) that the
mean z-score for the group significantly differs from 0
Questions:
1: Would I then need to somehow convert this what I think is more like a
t-map back to a z-map somehow in order to perform a cluster analysis
which only "works" on z-values?
2:Do I just use an individual person's z-map to estimate spatial
smoothness as I thought was suggested by Andrew Holmes in
http://www.mailbase.ac.uk/lists/spm/1998-07/0034.html ?
3:4: Details of t-test approach (Please check if possible): To perform a
t-test on the z-scores at a Talairach point, I find the following group
"z-score" using a one sample t-test, that is the group
"z-score" =
(mean z-score for group)/ standard error =
(mean z-score for group)/ sqrt (variance/7)
where variance = [ (subject1's z-score - mean z-score)^2 + (subject2's
z-score - mean z-score)^2 +...+ (subject7's z-score - mean z-score)^2 ]/6
I would then have to convert the resulting group "t-map" into a "z-map"
somehow.
4: Is there any problem with my approach?
Thanks alot
Jeff Lorberbaum
PS: I use spm96 as part of medx3.3.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|