Hi Adrian,

You may want to compute the F-test below which will tells you whether at least one of your condition differ from all other included in your model. It is a standard 2nd level F-test contrast for 1way within subjects ANOVAs.

ncond = 4; nsubj = 11;

F = eye(ncond)*(ncond-1)/ncond + (ones(ncond)-eye(ncond))*-1/ncond;
F = [zeros(ncond , nsubj) F];

It assumes as in your case that the factor subject in your ANOVA precedes the factor condition.

The first expression calculates the following matrix :

 0.75  -0.25   -0.25   -0.25
-0.25   0.75   -0.25   -0.25
-0.25  -0.25    0.75   -0.25
-0.25  -0.25   -0.25    0.75

Best wishes

Swann



2010/11/25 Jonathan Peelle <[log in to unmask]>
Hi Adrian

> 1. Are the above contrasts correct.

Yes, they seem fine.


> 2. Why do the contrasts have to sum up to 0? I thought the F-contrast should be [zeros(4, 11) eye(4)], but SPM says "invalid contrast".

When you have columns for subject effects, it's not possible to
directly estimate the mean for certain conditions.  I.e., if you look
at the indicators along the bottom of the design matrix, they will be
gray.  This means you can't do the equivalent of a "[1 0...]" contrast
with these columns, and thus an eye(4) contrast is similarly invalid
(because each row contains an invalid contrast).  Comparing conditions
is still fine (as you've discovered).


> 3. Can I simply multiply my contrasts by -1 to test for "negative" effects?

Yes.

Hope this helps!

Jonathan



--
Swann Pichon, PhD
Laboratory for Behavioral Neurology and Imaging of Cognition
Department of Neuroscience, University Medical Center
1 rue Michel-Servet, 1211 GENEVA 4, Switzerland
Tel: +41 (0)22 379 5979
Fax: +41 (0)22 379 5402
Gsm: +33 (0)6 26 43 83 61
http://labnic.unige.ch/