Dear spm users,
I will first shortly explain my paradigm in order to be understood;
I made up a study with 6 subjects wich were required to performed one
task.
This task, let's called it "U", was experimentally modulated with 2
differents factors :
Factor 1 : i [ 0 <= i <= 3 ]
Factor 2 : j [ 0<= j <= 2 ]
In a first analysis, I defined( U(i,j) [hrf + d/dt hrf] x Runs x
subjects) regressors (1452 + 30 beta !) in order to question SPM for
linear (soubstration) contrasts:
for instance : U(3,0) minus U(0,0);
in this case the constrast used was Cx = repmat( [ -1 - 1 0 ... 0 1 1
] , 1 , number_of_subjects);
As I wanted to tested the linear effects of i (and j) on U, I also
performed the following contrasts :
Cx = repmat([ 1 1 2 2 3 3 4 4] - mean([1 1 2 2 3 3 4 4]), 1,
number_of_subject);
Explanation = [ 1 1 .. = ponderation of +1 and +1 for hrf (
U(1,j)) and d/dt(hrf) (U(1,j)
The "mean" substraction was introduced in order to "deleted" the commun
effects of i on U
This contrats seems to show the linear effect of i on U;
But when I tried this method with the parameter j : for instance :
Cx = repmat([ 1 1 2 2 3 3 ] - mean( [1 1 2 2 3 3] ), 1,
number_of_subject);
in this case :
([ 1 1 2 2 3 3 ] - mean([1 1 2 2 3 3]) = [ -1 -1 0 0 1 1 ]
So the effects of j = 2 is not taken in account in the analysis ( [ ...
0 0 ... ] ) !
My question is ... : how may I questioned SPM99's-contrast in order to
test linear (or other) effects of i and j factors on U ?
If explaintions and my question are not very clear (...and they are, I
guess ...) please mail me for additionnal details.
Thank U in advance for your help ;
J.B.
*************************
J-B Pochon
14 Bd Ornano
75018 Paris - France -
[log in to unmask] or
[log in to unmask]
*************************
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|