2009/10/17 Lin Yun
<[log in to unmask]>
Dear SPMers:
I'm a rookie on SPM, and currently doing a block designed fMRI
experiment. I have 15 subjects, each performed 3 identical runs of my test,
after 1st level analysis, gave me 45 contrast maps.
My question is:
(1) Will the contrast map be different if I applied
different statistical level during the first level analysis?
what statistical level are you referring to? you should not be thresholding the first level con images, but if you did it would only change the result by eliminating sub-threhold activation.
(2) How should I be doing the 2nd level analysis, taking all
45 contrast map into account, or discard those with undesired results? Some
collegues in our department suggest me to discard those with
undesired results, although no other interference (i.e movement) was
found.
Each subject should only have 1 contrast image for each covariate of interest, even if it occurs in more than 1 session. Redo the analysis to combine betas from each session into 1 con.
As for discarding contrasts with undesired results, perhaps this is just a semantic mistake but that is horribly bad science depending on your definition of undesired. Why don't you descibe how you would systematically classify a contrast map as undesired. If you have just cause to remove a subject, you can leave them out of 2nd level stats, but obviously you can't just choose what to take up to the 2nd level based on what fits your hypothesis.
(3) I conclude all the 45 contrast maps, it really gave me a
glass brain with barely activations. How can I fix that?
what test did you do? was this result corrected?
Thanks for your time, any suggestion is
appreciated.
Yun Lin
Graduate student
Department of radiology
Southwest hospital, 3rd Military medical university,
China.
