>
Dear Shan,
Thanks for your email. I'm replying on behalf of Karl.
There are basically two ways to analyse your data.
The first is a 'fixed-effects (FFX) analysis' which you can implement
in SPM using the 'multi-subject:conditions and covariates'
design option. This makes a single
model into which data from all the subjects is entered at the same time.
After telling SPM which images are associated with which
subject and which condition (rest or stimulation) SPM will then
compute the parameters of a general linear model.
Using the RESULTS option you can then set up a 't-contrast'
which allows you to see the positions in the
brain for which activity is significantly greater in the
stimulation condition than in the rest condition.
Your inference from this is 'there is a significant increase in activation in
positions X, Y and Z in the particular nine subjects
I scanned'.
The second way to analyse your data is to do a
'random-effects (RFX) analysis'. You can implement this in
SPM in two stages. In the first stage, you make
separate models and 't-contrasts' for
each of your subjects. This leads to a contrast image ('con*.img') for
each subject. In the second stage you put all
the 'con*.img' files into a single model. In SPM
you choose the 'one-sample t-test' design option. This tells you the
the positions in the brain where the average contrast is
significantly non-zero. Your inference from this is
'there is a significant increase in activation in positions X, Y and Z in
my population of subjects'.
So, the difference between the two methods is that with RFX you
can make inferences about the population from which the 9 subjects
came from, whereas with FFX you're making inferences about
the particular 9 subjects.
It is recommended you have at least 12 subjects to do a RFX analysis,
so in your case I'd do FFX.
To find out more about FFX see
http://www.fil.ion.ucl.ac.uk/spm/course/notes97/Ch3.pdf
and for RFX see the section in
http://www.mrc-cbu.cam.ac.uk/Imaging/spm_theory.htm
All the best,
Will.
> >From [log in to unmask] Wed Mar 28 06:46:03 2001
> Date: Wed, 28 Mar 2001 13:54:55 +0800 (CST)
> From: Shan Baoci <[log in to unmask]>
> To: karl <[log in to unmask]>
> Subject: Asking quistion
>
> Mr. Friston
> I am a new SPM user. I have a set of block design data of 9 subjects, the
> condition alternated between rest and auditory stimulation, each subject 4
> blocks, each block has a 20s rest and a 40s simulation. I want to know the
> active brain regions responding to stimulation. I have analysed each
> subject's data respectively according to the example, get the active map
> of each subject. I want to get the average active map of 9 subjects now,
> how should I do?
> I have done an processing in which the data of 9 subjects are treated as 9
> session(when SPM asking number of sessions, I select 9),the next steps are
> same as a single subject, the contrast that I used is 1 1 1 1 1 1 1 1 1.At
> the end, I get a active map. If this map is the average active map of 9
> subjects?
> Best wishes
> Baoci Shan
>
> ----- End Included Message -----
--
William D. Penny
Wellcome Department of Cognitive Neurology
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7478
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|