 |
 |
Dear all, I am currently analyzing data from a parametric fMRI design. Let's say that in my first level model I have 3 regressors corresponding to the same event (viewing a picture) but in three different conditions (viewing the picture in 3 different contexts). At each trial I also have a value that I enter as a parametric modulator (pmod). On the second level I focus on the parametric modulations only. That is why I build a within subject ANOVA with one factor (condition) with three modalities: Cond1*Pmod Cond2*Pmod Cond3*Pmod For the sake of clarity let's say that I focus on Cond 3. My problem arises here because I want to identify voxels that both correlate with the pmod above baseline (0 0 1) AND whose correlation differ from the other two conditions (-1 0 1 ; 0 -1 1). What would be the best way to implement that in SPM? Currently I came out with 3 different ways and I cannot decide which one would be the best: 1. Do a simple parametric modulation (0 0 1) with a threshold of p<.05(FWE) and identify significant clusters. Then go back to the first level to extract individual betas from each participant for the maximum voxel of each cluster (or the mean for the whole cluster?), for each condition. Then treat these beta values like any dependent variables and test them for differences with ANOVAs or t-tests. 2. Do a simple parametric modulation (0 0 1) with a threshold of p<.05(FWE) and then save the clusters as a mask. Do a conjunction between (-1 0 1) AND (0 -1 1) with an uncorrected p<.001 and save the clusters as a mask. Finally do a superposition of the two masks to identify common voxels. 3. Do a conjunction between (0 0 1) AND (-1 0 1) AND (0 -1 1). In this case which threshold should I use? Thank you in advance for your help. Matias Baltazar PhD Student - Teacher Assistant in Neuropsychology Paris 8 University 2 rue de la Liberte 93526 Saint Denis Tel : +33 (0) 1494 07108 [log in to unmask]