Dear SPM
I have a few questions regarding Bayesian analysis.
1) It seems that for bayesian second level analyses one has to set up the
contrasts BEFORE running the Bayesian estimation step. Is this correct and
why is this necessary?
2) The spm5 implementation seems to produce reasonable 'P' maps (reasonable
= not every voxel activated) when it converts F to P contrasts. However,
T-contrasts seem to show activation at every voxel in the brain. Is this due
to some obvious mistake? (I realize this is a difficult question to answer
since I haven't specified what I did other than to make a T-contrast, which
does not activate every or even most voxels in a conventional SPM analysis,
but perhaps the result suggests some basic error.)
3) Prior messages on the list suggest that it is possible to threshold the
PPM's by effect size. However, I can find no such option when clicking
results. One example email is from Nov 5, 2005 by Will Penny who says:
In SPM's Bayesian inference you can produce posterior probability maps
of the population
effect being greater than a user-specified size eg. 1% of the global
mean. In SPM's
Bayesian inference you can produce posterior probability maps of the
population effect
being greater than a user-specified size eg. 1% of the global mean.
Is this still possible, or am I misunderstanding this choice? Also in the
papers describing Bayesian analyses, figures that show the SPM interface
show a text label just below the MIP's that says "effect size: 0.X%" for
example. I don't see this in the current interface whether or not I select
plot effect-size in the questions for displaying results.
4) There was an email to the list from Will Penny I believe and I may have
gotten this wrong but it said something like Bayesian PPM's look at the
variance across voxels, and then seemed to imply that the 2nd level analyses
also did not account for intrasubject variance. Sorry if I got this wrong,
but is this true? Would a 2nd level Bayesian PPM across subjects. not
constitute a random effects analysis.
Thanks
Darren
|