Hi Jesper et al:
Many thanks for your response regarding the second level with
covariates. Your answer - do the first level and then enter the con
images from the individual covariate analyses into the basic t-test
model - was essentially my approach. Tom Nichols suggested, however,
that this procedure might not be correct because (and I am risking
misquoting Tom here so these comments are really mine not his :)) a
covariate analysis is inherently unbalanced. There is no balancing of
conditions because there are no conditions... Tom should probably
weigh in here as I am certain he will have a more sophisticated or
just plain correct query.
With regard to the two sample t-test, I was implementing it
incorrectly and thank Dave McGonigle for pointing me in the right
direction.
Now a new problem. I have a very simply designed fMRI study
consisting of four blocks of stimulation at varying intensities. The
blocks consist of stimulation and rest and I am able to analyse each
block comparing stimulus with rest without a problem. What I would
now like to do, however, is ignore the rest periods and compare
stimuli. How do I do this? Note that what I want to do is directly
compare the stimuli and not compare the increases of stimuli A
compared with rest versus stimuli B compared with rest. This latter
comparison is very easy to implement, i.e., I can do it, but the
direct comparison of stimulus A with B is proving much more
difficult, i.e., I cannot do it.
So far I have tried eliminating all the rest scans and assigning the
remaining scans, "PET like", to conditions A, B, C, D, but have had
no success.
Best wishes,
Stuart.
UPMC MR Research Center
PUH B-804
200 Lothrop Street
Pittsburgh
PA 15213
Sent by Medscape Mail: FREE Portable E-mail for Professionals on the Move
http://www.medscape.com
|