Amit Anand wrote:
> Will,
>
> What should we use for contrast for this simple correlation activation
> analysis. We have a box-car design with negative and neutral pictures. We
> have put the regressor at the ACC timeseries. if we use a contrast 1 -1 1
> then we do not get any difference from 1 -1. If we do 0 0 1 then ACC lights
> up but so does the rest of the brain.
A t-contrast of [0 0 1] is the correct contrast (assuming ACC is the third regressor) to look
for positive correlations between ACC and other voxels. What threshold did you use ?
eg. p<0.001 uncorrected ? (Note to look for positive or negative correlations
is the F-contrast [0 0 1])
We do not have another psychological
> variable such as attention so I do not think we can use PPI.
> Could you tell us what kind of contrast we need to choose to get a
> meaningful correlation between the ACC time series and other regions of the
> brain which also respond to the negative vs. neutral conditions in a box
> car design.
>
If you wished to look for voxels that changed their
correlation with ACC as a function of emotial content, E (defined as
negative-neutral pictures), then you could do a PPI.
Set up a VOI containing ACC (which I guess you've already done), by
pressing the VOI button.
Then press the PPI button. Your psychological variable is then
E (which you can extract using a [1 -1] contrast over the negative
and neutral picture regressors).
Then set up a new SPM analysis to look for PPIs. Follow the
instructions here for an example:
http://www.fil.ion.ucl.ac.uk/~wpenny/datasets/attention/README_GLM_PPI.txt
Best,
Will.
> Thanks,
>
> Amit
>
>
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7475
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|