Dear Volkmar,
>I think, the threshold is not really the problem. The crucial point is,
>that your results are based on completely different spm's, depending on
>which spm you select to be masked. Masking, as it is implemented in SPM99,
>is just an image manipulation, but no statistical operation. If you want
>to perform a statistics on 2 or more spm's to find common peaks, you
>should use conjunction.
I take your point that the two spm's will be different, after all this
is just masking of images, but the spatial distribution of the
inclusive masks should be the same. As I understand it the mask is
calculated from the intersection of the two images after statistical
thresholding and then applied to one of the images. For both of the
inclusive mask analyses (contrast1 masking contrast2 and contrast2
masking contrast1) to have the same spatial distribution they must
have the same thresholdings applied to them. I feel the method mentioned
in my prior e-mail allows this to be the case.
Yours,
Tom.
=================================================
Mr.Tom Barrick,
PhD Student,
Magnetic Resonance Image Analysis Research Centre,
University of Liverpool,
Pembroke Place, PO BOX 147,
Liverpool, L69 3BX.
Tel : +44 (0) 151 794 5634
Fax : +44 (0) 151 794 5635
email: [log in to unmask]
=================================================
|