Dear Hideki,
>I am a H215O-PET and SPM99 user.
>I have a simple question.
>
>I tried to use "Multisubj-covariate only" to find the brain regions correlated to the subjects' performance such as reaction time, but we
could't.
>We found why we could not. It seems that we have to choose at lease 2 images in one subject. When we chose one image in one
subject,
>Error message was displied.
>I want to choose one image in one subject in order to find the brain regions correlated to subjects' reaction time.
>Please let me know how I should do.
This is because block effects can't be estimated in your approach.
In the "Multisubj-covariate only"PET design, you need to
model and estimate (and factor out) the block effect for each subjects,
i.e. subject's mean response.
As we take 1 df for block effect, if there are only one image per one subject,
there are no leaving df for statistics as follows:
Df = (10 images) - (10 block effects) - (1 RT regressor) < 0 !.
That is why you need at lease 2 images in one subject.
If you want to investgate the correlation between RT and
"delta" rCBF in a brain region over whole brain, and
you have only one RT score per subject,
following approach may be took:
If your expemriment have "rest scans",
you make a contrast image (delta rCBF image) of task(s) vs rest(s) for each subject
and then enter it and RTs into "Simple Regression" in "Basic models".
Regards, Takanori.
-------------------------------------------------------------
Takanori Kochiyama
Department of Intelligent Mechanical Systems Engineering
Kagawa Univ., Hayashi-cho 2217-20,Takamatsu, JAPAN
Phone: +81-87-864-2337,Fax: +81-87-864-2369
e-mail: [log in to unmask]
-------------------------------------------------------------
|