A question that has come up a number of times concerns how to compute an
SPM that will contain the areas that are jointly activated in two random
effects analyses.
Based on the previous tip from Christian Buechel. I've conducted an
analysis that uses Imcalc masking to conduct this analysis. Hopefully
someone will correct this if it is in error.
1. conduct the two rfx analyses and use the "write filtered" button to
write out the result images for each analysis at the desired statistical
threshold. Note that the simple spmT images are not thresholded, you need
to use the write filtered option here.
2. Using Imcalc, mask one result image by the other, ( i1.*(i2>0)),
writing out the result.
3. Use Rik Henson's "dummy contrast" trick to overlay the resulting masked
image over a canonical T1 brain. At this stage the threshold should
presumably be p<.999, as the thresholding has already been effected by
step 2.
A remaining issue: what p value should one threshold the two rfx
contrasts at to get a reasonable joint p value for the overlap?
Stephan
_____
Subject: Re: Parametric modulation and conjunction in second level
analysis
From: Christian Buechel ([log in to unmask])
Date: 13 Dec 1999 - 12:12 GMT
>[question from SPM list]
> One more question: I am interested in the brain areas, that are
> activated in both of two conditions [] I thought about masking and
> conjunction. What would be more appropiate and how can I implement this
> in the second level analysis.
This is more complicated because conjunctions are not possible at the RFX
level. [.]
Masking is possible, however, your mask would stem from a separate RFX
analysis and SPM won't let you do that. However you can use ImCalc to mask
the images manually. E.g. i1 > 0. Simply write the result images at the
desired threshold and use ImCalc to create a new, masked image.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|