Dear Jubin,
>
>Dear spm list,
>we have acquired a dataset with fMRI and would now be interested in doing
>a statistical analysis on a between groups and multiple conditions basis.
>Briefly, we have two groups (5 subjects each) who performed each two
>different tasks (each task paired with an associated rest baseline
>condition). Would the following approach be correct?
>1) Perform a 1st-level single subject analysis with two conditions x
>subject (do the baseline scans have to be specified in this case?).
Correct, though you may want to consider performing a fixed-effects group
analysis at the first level since that will allow you to play about with
conjunctions across subjects as well. The results would be otherwise
identical.
>2) Perform a 2-nd level random effect analysis (Full monty design, 2
>groups, 2 conditions per group).
>
Not really. It would be better if you decided beforehand what
condition-by-group effects you were interested in and created the
appropriate first-level contrasts using the contrast manager, yielding one
con*.img per subject and contrast. You would then enter these at a second
level (one image per subject for any given subject-by-condition
interaction) as a two sample t-test.
>In addition, we would be interested in doing a principal component
>analysis on the same dataset. Is there anything as the <Eigenimage> option
>in SPM96 available for SPM99, yet?
Eigenimage analysis will not be part of SPM99, but rather of a special
toolbox for multivariate stats, the release date of which I am not sure.
>And if there is one, could a principal component analysis be applied on
>the design specified above (the goal is mainly to test orthogonality of
>the variance between the two groups)?
I am afraid I don't quite understand this one.
>
>Any comment would be greatly appreciated.
>Best wishes,
>Jubin Abutalebi
>
Good luck Jesper
Jesper Andersson
Wellcome Dept. of Cognitive Neurology
12 Queen Square
London WC1N 3BG
phone: 44 171 833 7484
fax: 44 171 813 1420
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|