Dear all,
I agree with Russ - what he says sounds intuitively true, and I have often
wondered myself whether the compression would make any difference in
practice. And, again intuitively, it seems that in most cases it
shouldn't make much difference.
Still, I wonder if it would be worth it trying to solve this problem in
principle, as was already suggested: i.e., find a way to compress and
uncompress the data (by changing spm_append.m) so that it doesn't lose any
precision in the process of doing so, but the size of the Y.mad file
doesn't increase either.
Is this possible at all? I wonder if other people with more experience
with data formats and compression would have some ideas.
Kalina
> hi all - in theory this data compression issue is a problem, but in
> practice
> I think that it's not really to be worried about if you are averaging
> over
> voxels. I think the main worry would come about in computing variance
> measures over data extracted directly from Y.mad from a single voxel. If
> others on the list think this is wrong please let me know.
>
> with the size of ROIs that we usually use (10-30 voxels) any quantization
> noise should be averaged out, and that noise should only impinge on your
> signal if you are working with very tiny effect. In practice we have
> never found this
> to be the case.
>
> cheers
> russ
>
>
>
> ----
> Russell A. Poldrack, Ph.D.
> UCLA Department of Psychology
> Franz Hall, Box 951563
> Los Angeles, CA 90095-1563
>
> email: [log in to unmask]
> phone: 310-794-1224
> fax: 310-206-5895
> web: http://www.poldracklab.org/
>
_________________________________________
Kalina Christoff, Ph.D.
MRC Cognition and Brain Sciences Unit
15 Chaucer Road
Cambridge, CB2 2EF, U.K.
Tel: +44 (0) 1223 355294 Ext 723
Fax: +44 (0) 1223 359062
Email: [log in to unmask]
_________________________________________
|