You're right that this design isn't accounting for repeated measures,
and has the wrong degrees of freedom. See the paired design examples
in the FEAT manual to see how to model out subject means and get the
dof correct.
--------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington,
Oxford. OX3 9 DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask]
http://www.fmrib.ox.ac.uk/~steve
----------------------
On 20 Feb 2009, at 22:56, David Shirinyan <[log in to unmask]> wrote:
> Lets say, for simplicity’s sake we have a study with 3 subjects and
> 3 within
> subjects conditions. We made 3 contrasts (corresponding to our 3
> conditions). Participants gave pain ratings during our study which
> we now
> want to correlate with the signal in the 3 conditions/contrasts.
> Before moving forward with the analyses, I wanted to run this by the
> group.
> I was thinking of running a higher-level analysis with 9 COPE files
> as inputs
>
> Input
> 1 Sub1COPE1
> 2 Sub2COPE1
> 3 Sub3COPE1
> 4 Sub1COPE2
> 5 Sub2COPE2
> 6 Sub3COPE2
> 7 Sub1COPE3
> 8 Sub2COPE3
> 9 Sub3COPE3
>
> I would model the data as follows
>
> Group EV1(condition1) EV2 (condition2) EV3 (condition3) EV4 (pain
> ratings)
> 1 1 0 0 3
> 1 1 0 0 4
> 1 1 0 0 7
> 1 0 1 0 4
> 1 0 1 0 6
> 1 0 1 0 9
> 1 0 0 1 7
> 1 0 0 1 5
> 1 0 0 1 10
>
> Orthogonalize EV4 with 1, 2, and 3
>
> Contrast for Positive correlation would be
> 0 0 0 1
>
> Contrast for Negative correlation would be
> 0 0 0 -1
>
> Is this correct? Should I instead model pain ratings per condition as
> separate EV’s then combine them as a contrast? Should my EV’s
> instead model
> a given participant? My main concern is in accounting for the fact
> that the
> 9 datasets are not independent.
>
> Many Thanks
>
|