Dear SPM users,
I'm performing an EEG-informed fMRI analysis on 11 subjects.
At first-level analysis, I have 5 EEG-derived time series representing the modulation of the alpha, beta, gamma, delta, theta rhythms. I convolved each EEG-derived regressor with canonical HRF and its time and dispersion derivatives. This resulted in 15 (5 x 3) regressors per subject and, consequently, in 15 t-contrast map per subject. I'm only interested in the effect of the single rhythm separately (i.e I want to test only whether the rhythms are different from zero and not whether they are different from each other).
In this case, which of the following second-level analysis is correct?
1) I define 5 (one per rhythm) different full factorial design with 1 factor (basis function) having 3 levels (canonical, time derivative, dispersion derivative). In this case, if, for instance, I would like to test for the effect of the alpha rhythm as characterized by the informed basis set, I should create a second-level design matrix including only the first-level t-contrast maps relative to the alpha rhythm and define a second-level F-contrast [1 0 0 ; 0 1 0 ; 0 0 1]. Similarly, to test the effect of the beta rhythm, I should create another second-level design matrix different from the alpha one and define again the second-level F-contrast [1 0 0 ; 0 1 0 ; 0 0 1].
2) I define a unique full factorial design with 2 factor (rhythm and basis function), the first factor (rhythm) having 5 levels (alpha , beta, gamma, delta and theta) and the second factor (basis function) having 3 levels (canonical, time derivative, dispersion derivative). In this case, if, for instance, I would like to test for the effect of the alpha rhythm as characterized by the informed basis set, I should have a second-level design matrix including all the first-level t-contrast maps relative to all the rhythms and define a second-level F-contrast
[1 0 0 0 0 0 0 0 0 0 0 0 0 0 0;
0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 ;
0 0 1 0 0 0 0 0 0 0 0 0 0 0 0].
Similarly, to test for beta effect, I should define the second level F-contrast
[0 0 0 1 0 0 0 0 0 0 0 0 0 0 0;
0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 ;
0 0 0 0 0 1 0 0 0 0 0 0 0 0 0].
In summary, my question is: should I include in the second-level analysis all the t-contrast maps obtained at the first level and perform a factorial analysis with two factor or can I select only some of them (specifically only one rhythm) and perform a one factor analysis? And if the two approaches are equally valid, what are the differences between them?
Thank you very much in advance,
Roberta
|