Hi
Yes, the F-test is calculated on the ICA time course and the estimated best model fit, just in order to give some indication if a particular component is within the space spanned by the design or not, i.e. can validly be called task-related. The test itself simply looks at the amount of variance in the ICA time course explained by the design, i.e. performs a simple variance-ratio test.
hth
Christian
On 17 Jun 2010, at 09:43, Silvia Storti wrote:
> Hi all,
> I'm Silvia Storti, PhD student of the University of Verona (Italy).
> I'm learning to use MELODIC toolbox, but I have a problem.
> In a single session ICA (200 volumes) I specified in Post Stats the temporal design matrix (2 columns) and the contrast matrix. I found that a component is task related (F-test on full model fit = 44.06, dof1 = 2; dof2 = 197 p < 0.00000 (uncorrected for #comp.)) but I don't understand how the F-test was perfomed and how these values were calculated. Was it performed on the estimated time course and on the model fit? Only on these two signals?
>
> Thanks,
> Silvia
|