Dear Joe and SPM gurus;
Thank you for your practical suggestion.
I choose the A+B-2C masked with A-C and B-C rather than
conjunction of (A-C) & (B-C) for the analysis.
I have searched the mail base for "conjunction", but failed
to find the answer for my concern.
Let me write my questions down again:
1) When one puts non-orthogonal contrasts in the conjunction
analysis in SPM96, what is violated?
2) Are the results created by the non-orthogonal contrasts
in SPM96 conjunction analysis completely nonsence or interpretable?
Thanks.
Michiru,
At 5:44 AM -0800 0.11.13, Joseph T. Devlin wrote:
>Hi Makuuchi-san,
>
>> I am hoping to find a conjunctional activation in an experiment
>> which has three conditions: A, B, and C.
>> My concern is that the conjuncional activation of (A-C) & (B-C)
>> can be obtained, since the two contrasts were not orthogonal.
>> In SPM96, this combination was accepted, but not in SPM99.
>> I would like to know whether it is completely wrong or nonsense
>> to perform this analysis in SPM96.
>
>SPM99 changes the contrasts in a conjunction of A-C and B-C because
>the original two contrasts are not orthognal -- which is a requirement
>of a conjunction analysis. To answer the question, "What areas
>are activated in common for the two contrasts (A-C & B-C)" you'll
>need to use masking. If you compute the main effect of A+B-2C
>and mask it with A-C and B-C inclusively, then you will get the
>areas activated by A & B (relative to C) and exclude areas which were
>not active in both A-C and B-C. The level of your inclusive masking
>will determine how conservative your test is. An inclusive mask of
>0.001 means that the effects are present in each study with at least that
>significance level.
>
>SPM96 computed conjunctions in a different (and less valid) fashion
>which allowed non-orthogonal contrasts. Even so, you still had to use
>masking in SPM96 to ensure the effect was present in both (or in all)
>contrasts.
>
>There are many emails in the archives about this topic but I'm afraid
>I don't have them at hand. Search for "conjunction" and spm96/spm99.
>
>Good luck!
>
> - Joe
+++++++++++++++++++++++++++++++++++++
Michiru MAKUUCHI
幕内 充
Cognitive Neuroscience,
Faculty of Medicine, University of Tokyo.
7-3-1 Hongo, Bunkuo-ku, Tokyo, Japan.
phone; +0081-3-5841-3690
fax; +0081-3-5841-3335
e-mail; [log in to unmask]
+++++++++++++++++++++++++++++++++++++
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|