Beginning to get the answer.
Just as a reminder, I have first and last points of each of my sessions greatly offset from the rest of the data.
I found the post below.
But still have a question in response to the question asked by Guillaume in that post.
I am using an F contrast that encompasses all the regressors I use to build my contrast of interest.
i.e. if I'm interested in a T contrast
[0 0 0 1 -.5 -.5 0 0]
I adjust the data with a F contrast
[0 0 0 1 1 1 0 0]
Does the null space of that contrast contain the offset term? Is this like mean adjusting my data?
or should it be
[0 0 0 1 .5 .5 0 0]
or does that not matter?
Thanks a lot,
Max
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=SPM;VQ5Uzg;20100215184029%2B0000
Dear Jason,
see the help text in spm_regions.m:
what is displayed is the first eigenvariate of the filtered and adjusted
response in all suprathreshold voxels within a specified VOI. In
particular, the filtering consists of a temporal high pass filter and
whitening (non-sphericity correction), see l. 134-135, while the
adjustment is made by specifying a bi-partition of the design matrix
through an F-contrast, see l. 151.
Which F-contrast did you use to adjust the data? Does the null space of
that contrast contain the offset term (effectively mean-centering the
data)? I currently manage to reproduce what you observe only if I don't
adjust the data with respect to its mean.
Best regards,
Guillaume.
Jason Stretton wrote:
> can anyone explain what the graph represents when clicking the
> eigenvariate button? I'm particularly confused by the first and last
> response being substantially higher than the rest of the time-series?
>
> Thanks
>
> Jason
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|