Dear Joanna,
> I have four conditions -- 4 is the baseline. When I parametrically
> modulate the first three conditions with linear "other", learning scores,
> I end up with a 7 column design matrix. like this
> cond1 cond1PM cond2 cond2PM cond3 cond3PM baseline
>
> In order to see areas of activation in condition 1 that are
> linearly related to learning scores do I use the contrast
>
> 0 1 0 0 0 0 0
>
> or
>
> 0 1 0 0 0 0 -1 ?
>
Use the first one. The second column (cond1PM) is representing the slope of
the relationship between learning scores and evoked activity, so it doesn't
make sense to compare the slope of this relationship with baseline (which
doesn't have a slope, if you see what I mean!). The relevant contrast for
whether there is greater mean activity in cond1 relative to baseline is [1 0
0 0 0 0 -1]. You might like to mask your first contrast with this
cond1/baseline contrast if you want to restrict your examination for a
linear component to those areas which show cond1 activation over baseline
overall.
The first contrast ([0 1 0 0 0 0 0]) will identify areas where the slope of
the relationship between the condition 1 parameter and activity is
significantly different from zero. To test for different slopes between
conditions, you could use contrasts like [0 1 0 -1 0 0 0] and so on.
Note that the areas that you can identify with this design matrix are those
with a linear *component* to their relationship with the parameter. You
shouldn't conclude that such areas show a linear (as opposed to nonlinear)
relationship overall without also modelling higher order terms and showing
that the model fit is not improved by adding such terms.
best wishes,
Geraint
--
Dr. Geraint Rees
Wellcome Advanced Fellow, Lecturer,
Division of Biology 139-74, Institute of Neurology,
California Institute of Technology, University College London,
Pasadena CA 91125 London WC1N 3BG
voice 626-395-2880 020-7833-7472
fax 626-796-8876 020-7813-1420
web http://www.klab.caltech.edu/~geraint
--
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|