Dear Saaussan,
> I'm doing fmri on cervical spinal cord (not sure yet how important it
> might be clinically).With a simple on/off one hand clenching
> paradigm. I have the following problems in data analysis. Because of
> the small cross sectional area, i can't apply an exact mask on my data
> (risking that smoothing will wipe out some information). This makes
> it difficult to set a criteria for corrected p, because spm uses the
> pooled variance in calculating the T map.
In fact SPM uses voxel-specifc estimates of error variance (that's one of its
defining features).
> In the axial section i have
> 79 resels (not perfect masking) in coronal and sagittal sections it is
> 150 resels) the degrees of freedom is around 30.
>
> As far as i understand, a large volume is a criteria that needs to be
> met for the gaussian field theory used in estimating p. my question is
> how large this volume should be?
Generally an order of magnitude larger that tbe FHWM will guarantee
valid results. Keith's unified theory (implemented in SPM99) can now
deal with any [small] volume. It sounds as if your search volume is
sufficiently large (if not too large given you know where you are
looking).
With best wishes - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|