Hi. A few points here:
When you feed first-level FEAT results up to second level analyses it is
not at all necessary to have equal number of time points in the
first-level sessions. There is no bias in comparisons if the number of
time points varies.
Whether you have the same number of time points or different numbers, it
is VERY important to keep the "height" of the EVs the same in all
sessions. Otherwise the estimated effect sizes (PEs - parameter estimates,
and COPEs - contrasts of PEs) do not mean the same thing as each other and
cannot be sensibly combined at higher level.
Hope this all makes sense!
Thanks, Steve.
On Wed, 26 Feb 2003, John Herrington wrote:
> I would like to statistically compare fMRI data for two conditions (within
> a single subject/run) - the first condition corresponding to 84 fMRI
> images, and the second to 168 images. In other words, one condition was
> on for twice as long as the other. When setting up this analysis in FEAT,
> I am inclined to treat the data as if they are nothing more than two
> datasets with unequal Ns, and create EVs where each image in the first
> condition is coded with a 1, and each image in the second is coded with a
> .5. I can then contrast those two EVs, applying a weight of 1 to the
> first EV, and -1 to the second. My questions: is this how you would
> choose to set up such an analysis? If not, how would you do it? Are
> there any other considerations in implementing this type of analysis above
> and beyond what's inherent statistically in comparing data with unequal Ns
> - either within the context of FEAT, or in general for fMRI data?
>
> Thanks,
>
> John Herrington
>
Stephen M. Smith MA DPhil CEng MIEE
Associate Director, FMRIB and Analysis Research Coordinator
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|