From: Arturo Hernandez <[log in to unmask]>
>I am trying to understand exactly what the difference between masking and
>conjunctions is. If I have two contrasts (F1 vs. Rest), (F2 vs. Rest) what
>is the difference between the conjunction of these two contrasts and F1
>masked by F2.
From: Tugan Muftuler <[log in to unmask]> recently asked
> what does "masking" (and conjunction analysis) do?
SO:
SOME INFORMATION ON MASKING & CONJUNCTIONS
Say you have 2 contrasts
Contrast 1 = Activation 1 - Baseline 1
Contrast 2 = Activation 2 - Baseline 2
If you mask Contrast 1 with Contrast 2, SPM will display the voxels
significant in the main contrast (ie contrast 1) that are also significant
in the mask (ie Contrast 2). In other words, any voxels that are not
significant in the mask are masked out. Masking does not alter the
statistics associated with the main contrast. The thresholds can be altered
independently so for instance you could specify the mask at p<0.05 and the
main contrast at p<0.001.
MASKING therefore identifies regions that are activated in BOTH contrasts
but the Z scores relate to the statistics from the main contrast only.
In a CONJUNCTION analysis, you create a new contrast summing over two or
more contrasts. For instance, the conjunction of contrast 1 and contrast 2
is calculated from (Activation 1 + Activation 2) - (Baseline 1 and
Baseline2). The statistics displayed will therefore be different from
those in the individual contrasts. In addition to summing over 2 or more
contrasts, conjunction analysis removes any voxels where there is a
significant difference between the specified contrasts. In this sense,
there are similarities with the masking option.
Masking removes voxels that are not significant in both contrasts
Conjunction analyses removes voxels where there is a significant difference
between contrasts.
Masking does not create new contrasts or statistics,
Conjunction analysis does create new contrasts and statistics.
The use of masking or conjunction analyses depends on your question.
If you want to identify regions that are significant in contrast 1 and also
in contrast 2 at a particular threshold, you use masking.
Conjunction analysis is more sensitive ( and less conservative) and
attributes Z scores to the co-occurrence of activation. The manner with
which this is done is currently under review and will change in SPM99.
Over to Karl if you need more details.
With best wishes
Cathy Price
Wellcome Department of Cognitive Neurology
Institute of Neurology
12 Queen Square
London WC1N 3BG
Tel: 44 171 833 7455
Fax: 44 171 813 1445
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|