Hi Cyril and Miles,
just a quick clarification about contrast 2 -1 -1.
This compares the activation during Encoding to the *average* activation
of Maintenance and Recall.
and does NOT necessarily imply "this tells you which voxels are more
activated by encoding than maintenance and recall"!
Think of activation being 4 1 5, i.e. higher during Recall than both
Encoding and Maintenance, then the t-test with contrast [2 -1 -1] will
test for (2)*4 + (-1)*1 + (-1)*5 = 2 which could be significantly larger
than 0...
To ensure that activation during Encoding is truly larger than during
Maintenance and Recall, then you would need to employ a conjunction
between these contrasts: [1 -1 0] & [1 0 -1].
Best,
Chris
Cyril Pernet a écrit :
>
> Hi Miles
>
>
> > For a working memory task, what is the difference between these two
> > contrasts:
> > I am interested in the activation just associated with the encoding
> phase.
>
> both gives you activations related to encoding but ...
>
> > The three events are:
> > Encoding, Maintenance, Recall
> > 1 0 0
>
> this tells you which voxels are significantly more active than the
> mean/rest (simple effect)
>
> > 2 -1 -1
>
> and this tells you which voxels are more activated by encoding than
> maintenance and recall
>
> to make things more complicated, I would suggest an additional
> option: use the contrast 2 -1 -1
> masked (inclusive) by 1 0 0, i.e. you can look for voxels more
> activated by encodng than he other conditions but only in areas where
> encoding is significantly bigger than the mean. This has a double
> advantage: 1. conceptually it makes sense to consider an 'encoding'
> voxel as being more active for encoding than the other conditions
> while making sure it is 'active' on it's own and 2. you are avoiding
> negative voxels i.e. a voxels with signal values e.g. -1 for encoding
> and e.g. -3 for the other conditions is more active for encoding right
> ? but you are not really interested in that ... you only want voxels
> with positve values for encoding, hence the mask.
>
> hope this all makes sense to you
>
> Cyril
>
>
>
> --
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
|