Dear Michael,
> Please clarify why a two condition FIXED RESPONSE block design does not
> yield identical results when analyzed with a t-test and an F-test. Is
> it because the ideal model does not predict a perfect "box-car" shaped
> response curve (i.e., only zeros and ones)? If this is true, then what
> is the default function that the perfect boxcar is convolved with using
> the fixed response selection, and is there a way to creat a perfect
> box-car wave form?
The perfect box-car wave-form obtains from convolving the box-car with
the perfect HRF (hemodynamic response function). We are still looking
for the perfect HRF (but it changes from voxel to voxel so we use a
canonical approximation or expand the HRF in terms of basis
functions).
The reason T and F tests give different results is that the F test is
effectively two tailed and identifies activations plus deactivations.
For statistical models with one d.f. the F is the T value squared.
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|