Dear Yuhong,
Each of the con***.img files that you enter into the random effects
analysis contains one value at each voxel location. This value equals
(Beta weight for trial type D's regressor) - (Beta weight for
trial type C's regressor).
Assuming that you've modelled the amplitude of the response for trial types
D and C, the difference between the Beta weights reflects a difference in
the amplitude of the hemodynamic response produced by trial types D and C.
Assuming that your con* images were generated by contrasting trial type
D with trial type C for each subject, entering a '1' in the random effects
contrast manager indicates will test whether trial type D produced a greater
response than trial type C at every voxel in the image. Entering a '-1' in
the contrats manager would allow you to test whether trial type C produces a
greater response than trial type D at every voxel within the image.
In either case above, I believe that a one-sample t-test calculates a
t-value at every voxel in the brain using this formula:
(Mean Difference Between Beta Weights Across Subjects) / (Standard Error
of the Mean Difference Across Subjects)
Hope this helps!
:> Daniel
Daniel Weissman, PhD
Center for Cognitive Neuroscience
Duke University
Durham, NC 27705
phone: (919)-681-1029
fax: (919)-681-0815
e-mail: [log in to unmask]
----- Original Message -----
From: "Yuhong" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Wednesday, July 25, 2001 6:43 PM
Subject: random effect: how to set up contrast
> Dear SPM group,
>
> I just started to do random contrast. Here is my design:
>
> 12 subjects, four conditions (A B C D). I'd like to find out whether D is
> significantly greater than C in a random effect analysis.
>
> I did the following:
> (1) For each subject, get a Con***.img and Con***.hdr for the contrast of
my
> interest (D-C).
> (2) In SPM99, choose "Basic Models" > "one-sample t-test", and enterred
each
> subject's Con***.img as the images. The computation created an output
> "SPM.mat".
>
> My question comes in the next step. The SPM program asks me to use
"Results"
> to assess results. So I clicked "Results", selected "SPM.mat" that I just
> created. But then the program requires me to specify a contrast. I am
> completely at a loss as what the "contrast" means. I tried some numbers,
and
> it seems to ask me to input one number (e.g., 1) rather than a true
contrast
> (e.g., 0 0 1 -1, with the sum = 0). Could you please explain what
"contrast"
> means in this step?
>
> Your help is greatly appreciated. Thank you.
>
> Yuhong
>
>
>
|