Some more "pooled variance" questions for Matthew Brett, Karl Petersson, and
other SPM statistical folks who care to comment:
1. What would be the correct approach in a two level (random effects)
analysis to using the Worsely pooled variance? a) "Worsleyize" the
individual subject contrast images before subjecting them to a random
effects (2nd level) analysis? or b) Worsleyize at the 2nd level.
Intuitively, a) seems to be the answer to me.
2. Does SPM99 have a way to determine if, in fact, variance is not
substantially different across the brain (thus allowing a valid pooled
variance approach)?
Thanks, all.
David Kareken
-----Original Message-----
From: Matthew Brett [mailto:[log in to unmask]]
Sent: Friday, August 25, 2000 12:17 PM
To: [log in to unmask]
Cc: Kareken, David A.; [log in to unmask];
[log in to unmask]
Subject: Re: Pooled variance III
Dear all,
I think the attached code does what you want. It uses the pooled variance
across voxels in the analysis to write new SPM_T images.
You need to have entered and evaluated your contrast using the SPM results
interface, generating the SPM T image. Thence, from the matlab window;
worsleyize
Select the analysis and contrast you want to convert, and it will write out
a
new image with the name spmWZ_<contrast no>.img. As you can see, the code
is
mostly interface, the meat of the thing is only a couple of lines.
Please let me know if there are any problems.
Best,
Matthew
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|