i used spm96 to perform the statistical analysis explained below: a four
condition conjunction analysis composed of two tasks and their respective
two baselines. i ran the analysis as a multistudy, multiple condition
contrast with the first study weighted -1 1, and the second study weighted
-1 1 as well. when i got to the results section, i clicked on SPM{Z}, then
answered yes to do a conjunction analysis. my question is this; what do i
enter into the "contrasts? 1 - 1" box?
nothing i do seems to work -i keep getting error messages. any help would
be greatly appreciated.
thanks.
-andre
>Dear Andre
>To find the conjunction of A1 and A2, all you need to do at the analysis
>stage is to specify the two contrasts seperately. eg
>
>A1 B1 A2 B2
> 1 -1 0 0
> 0 0 1 -1
>
>The rest of the analysis is all at the results stage. In SPM96 and 97,
>there is an option to do a conjunction analysis when you view the results.
>You respond Yes, and indicate the two contrasts. This analysis will sum
>up the two simple main effects and remove voxels where there is a
>significant difference between the contrasts. In SPM 97 you are given the
>option of eliminating voxels where there is a significant interaction
>between effects at a threshold of p< 0.05, 0.01 or 0.001. The more voxels
>you eliminate the greater the similarity between your 2 contrasts.
>Therefore, it is more conservative to remove voxels at 0.05 than 0.001.
>However, even when you eliminate voxels at 0.05, the conjunction analysis
>as implemented in SPM96 and 97 may still leave voxels where there is a
>significant effect for one contrast but not the other when the interaction
>is not significant at 0.05. To demonstrate that both contrasts are
>significant , it is therefore necessary to mask the conjunction with the 2
>contrasts (the next option in SPM). Again you can specify the threshold
>for the mask. A really impressive conjunction might survive a mask of
>0.001 for each contrast.
>
>If you do not have a version of SPM that calculates the conjuction, a
>simple alternative is to specify a contrast for the main effect (ie 1 -1 1
>-1) as well as the contrasts for the individual effects. Then mask the
>main effect with the individual effects. The issue here is that the main
>effect underestimates the significance of the results that are replicated
>in two independent contrasts. If the contrasts were totally orthogonal
>(which they are not), then a p value for the replication of 2 independent
>results each at 0.001 would be approximately 0.001 X 0.001.
>
>The conjunction analysis in the newest version of SPM is being adjusted to
>incorporate the masking procedure and generate the corrected p values.
>Meanwhile it is probably best to give the Z scores for the conjunction of
>the 2 contrasts (or the main effect) and for each contrast individually.
>
>With best wishes
>Cathy Price
>
>
>>We are conducting a study in which we have run four conditions - two
>>activation tasks (A1 and A2), each with its own baseline (B1 and B2).
>>
>>We would like to identify activations that are shared by these task pairs -
>>i.e. common processing differences as defined by Price and Friston in the
>>1997 paper "Cognitive Conjunction: A New Approach to Brain Activation
>>Experiments" published in Neuroimage.
>>
>>We need some practical detail - i.e. what SPM routines are to be used to
>>run such an analysis? How should the conditions and baselines be weighted?
>>
>>Andre Guillemin
>>National Institutes of Health, NIDCD
>>Bldg 10 Rm 3C716
>>9000 Rockville Pike
>>Bethesda, MD 20892
>>Phone: 301-435-5145
>>Fax: 301-480-7410
>
Andre Guillemin
National Institutes of Health, NIDCD
Bldg 10 Rm 3C716
9000 Rockville Pike
Bethesda, MD 20892
Phone: 301-435-5145
Fax: 301-480-7410
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|