Hi - I think you would find this cross-subject analysis a lot easier
if you treated this as a 3-level analysis, doing a separate fixed-
effects analysis for each subject at second level, that creates a
contrast for each A?B?-related contrast, and then you can do simple
group means and differences at the 3rd level for each second-level
(e.g. A1>A3) contrast.
Cheers.
On 3 Aug 2008, at 02:10, Heather L. Urry wrote:
> Hi there,
>
> I'm doing an across-subjects analysis of a mixed fMRI design. I just
> wanted to check on a couple of things and hope you can help.
>
> I've got n=15 in one group and n=16 in a second group, plus there are
> two repeated measures factors (one with 2 and one with 3 levels). What
> I'm planning on doing is modeling the repeated measures effects in the
> EVs as shown below with two example subjects in the last two EVs (if
> the columns don't line up properly, it's replicated in the attached
> txt file):
>
> EV1 EV2 EV3 EV4 EV5 EV6 EV7
> subj cope A1>A3 A2>A3 B2>B1 A1*B A2*B subj1 subj2...
> 1 A1B1 1 0 -1 -1 0 1 0
> 1 A1B2 1 0 1 1 0 1 0
> 1 A2B1 0 1 -1 0 -1 1 0
> 1 A2B2 0 1 1 0 1 1 0
> 1 A3B1 -1 -1 -1 1 1 1 0
> 1 A3B2 -1 -1 1 -1 -1 1 0
> 2 A1B1 1 0 -1 -1 0 0 1
> 2 A1B2 1 0 1 1 0 0 1
> 2 A2B1 0 1 -1 0 -1 0 1
> 2 A2B2 0 1 1 0 1 0 1
> 2 A3B1 -1 -1 -1 1 1 0 1
> 2 A3B2 -1 -1 1 -1 -1 0 1
> .
> .
> .
>
> Then, in order to get the group effects, I need to make contrasts
> using these subject EVs, right? So it would look like this (with only
> 2 subjects in EVs 6 and 7)):
>
> EV1 EV2 EV3 EV4 EV5 EV6 EV7
> grp1 > grp2 0 0 0 0 0 1 -1
> grp2 > grp1 0 0 0 0 0 -1 1
>
>
> And then to see if there are group differences in the repeated
> measures effects, something like this (for grp1>grp2 in EVs 6 and 7):
>
> A1>A3 1 0 0 0 0 1 -1
> A2>A3 0 1 0 0 0 1 -1
> B2>B1 0 0 1 0 0 1 -1
> A1*B 0 0 0 1 0 1 -1
> A2*B 0 0 0 0 1 1 -1
>
> Yes? No? Utter hogwash?
>
> Finally, I was thinking the group contrast should sum to zero, but it
> won't with different numbers in each group unless I compensate. So
> let's say there are 3 subjects total, 1 in grp1 and 2 in grp2. For
> grp1>grp2, would I enter [2 -1 -1] for the three subject EVs to
> compensate, or is this unnecessary?
>
> Cheers, and thanks in advance for your input,
>
> Heather
>
> --
> Heather L. Urry
> Department of Psychology
> Tufts University
> 490 Boston Avenue
> Medford, MA 02155
>
> email: [log in to unmask]
> phone: 617-627-3733
> fax: 617-627-3181
> <mixed_design.txt>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|