Sorry for multiple sends, I had a weird problem with my e-mail,
Dear SPMers,
- I have an er-fMRI study with two conditions of interest (A and B)
repeated in two different sessions :
(As session 1 : A + Cont., sess 2 : A + Cont., sess 3 : B + Cont. and sess
4 : B + Cont.)
For each of my 14 subjects, using SPM 99, I modelled the data using a
parametric model, in which I included paired sessions,
(1+2) and (3+4) specifying that there were "replicated conditions", with
different timing values.
So for each of my individual subjects I can obtain the main effect of the
different events in the different sessions using a simple t-test, leading
to con images for : A in sess1, A in sess 2, B in sess1 and B in sess 2.
Now, I would like to run a Random Effects analysis on these contrasts and
I'm mainly interested in establishing :
i) the activated regions for cond. A and B separately.
What kind of basic model shall I use ?
I guess a one sample t-test on the 28 .con images coming from two different
sessions wouldn't be correct, so I found on the mailing list the script to
perform a regression analysis with no constant term to get the conjunction
effect of A in session 1 and A in session 2.
Does this sound Ok ??
Is there another procedure that would be statistically more acceptable ?
ii) the regions activated both by A and by B.
Following the same idea I performed the same regression analysis (with no
constant term) on all the 56 .con images coming from the 4 sessions and
defining two main groups of data (one for A, one for B) but I'm not really
confident with the statistical value of this approach.
Is it ok to do the test like that or, is there another way to do it ?
Many thanks in advance,
Sincerelly,
Michel Hoen.
Institut des Sciences Cognitives
Equipe : "Cognition Séquentielle et Langage"
UMR 5015 CNRS-UCBL
67, Bd Pinel
69675 Bron Cedex
Tél : 04 37 91 12 65
|