Peter Clerinx wrote:
> Hello,
>
> I have a few (ok, I admit, a lot of) questions concerning contrasts in SPM. We have done a study
> on 25 subjects, with a block design of 2 conditions + baseline.
>
> When I do a analysis with a [1 0 1 0 ...] t-contrast vector to test cumulative effects of
> condition A, is this the same as [1/25 0 1/25 0 ...], or should I always normalise (sum to 1)
> these kind of contrasts ?
There's no need to sum to 1 here.
Multiplying (or dividing) a contrast vector by any aribtrary
number, scales both the numerator and denominator of the ensuing t-statistic
in the same way - so your t-values will be unchanged.
> Is this correct for a multiple subject analysis, or only appropriate
> for combined effects within the same subject ? Does this contrast mean I'm testing a H1:"at
> least one subject activates for condition A significantly" ?
>
A 'second-level' or 'random effects analysis' is an analysis whose inputs
are the con*.img's from a
'first-level' - or 'within-subject' analysis -
see
http://www.fil.ion.ucl.ac.uk/spm/doc/books/hbf2/pdfs/Ch12.pdf
for more info.
In this context, your inference is about the mean activation
level, m, in the population from which your subjects were drawn.
H0:m=0, H1:m not equal 0.
For, 'first-level' or 'within-subject' analysis the inference concerns
the effect, a, in that subject: H0:a=0, H1:a not equal to 0.
> When I'm using a differential contrast, e.g. [1 -1 1 -1 ...] for effect A > effect B, I get a lot
> of activation. Should I mask with the A>0 contrast ... ?
Yes - if you only want those voxels for which the response to A is positive, and the
response to B is even more positive.
This would filter out those voxels for which the response to A is negative (ie. a
deactivation - relative to mean activity in that voxel), and the response to B
is less negative.
BEst,
Will.
--
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/
|