Dear SPMers,
I repost this message, just in case it went forgotten...
I found the following old mail in the SPM mailing list. There Dr. Andrew
Holmes described a method ('interaction-of-contrasts' analysis) for doing
special between-group interactions. I applied this method to my data and I
would be happy if I could cite a reference addressing this problem.
Thank you in advance
my best regards
Luca Finelli
At 11:55 20.02.98 GMT, you wrote:
>Dear Dr Hellwig,
>
>| just recently I started to run analyses of PET-studies with SPM96b. I
>| don't have much technical and statistical background (being just a
>| trained linguist), so the help files for SPM are very difficult to
>| understand for me and I often have the feeling that I am lost, as they
>| are too cryptic.
>
>Fair point. The SPMcourse notes provide a more comprehensive and
accessible overview of the requisite theory, and are freely available. See:
> http://www.fil.ion.ucl.ac.uk/spm/course/notes.html
>
>Also, SPM96b was the public beta edition. There were a few bugs, and
>most people are now running current release version, SPM96. For details
>of the bugs, the patches, and how to upgrade, see:
> http://www.fil.ion.ucl.ac.uk/spm/spm96.html#Release
>
>
>| I assisted in a PET-study with 10 subjects. There were 6 different
>| conditions, and every subject got each of the six conditions once, making
>| 60 scans.
>|
>| I not only want to compare different conditions (I used Multi-subject:
>| different conditions and specified the contrasts, we were interested in),
>| but now I have also a between subject factor, gender, since 5 subjects
>| were male and 5 female. I don't know how to procede. How can I find out,
>| whether males and females show different activations?
>|
>| This is probably a question I should be able to find out myself, and I
>| have spent some time going through the SPM-archives in the hope to find
>| the answer this way, but I was not able to succeed.
>
>This has been discussed on the SPM list fairly recently: See
> http://www.mailbase.ac.uk/lists/spm/1997-12/0004.html (AS)
> http://www.mailbase.ac.uk/lists/spm/1997-12/0006.html (KJF)
> http://www.mailbase.ac.uk/lists/spm/1997-12/0013.html (APH)
>
>However, It's quite a complicated issue, surrounding which there's
>still a fair amount of confusion, so I'll briefly address your design
>here, and make a few general comments on the issues for wider
>consumption:
>
> ----------------
>
>You have five males, five females, each of which has six scans under
>*different* conditions. Since there are no replications of condition
>within subject, you can use a relatively simple model:
>
>* Use Multi-Study: different conditions
>* Enter males as one study, females as the other study,
> making sure the scans are entered in the same condition order (within
study).
>
>Say the conditions are B A1 A2 A3 A4 A5, entered in that order for each
>group, with B a baseline condition and A's the active conditions. SPM
>will fit condition effects seperately for each group, giving 12
>condition effects of interest (below, in the obvious notation)
> B(M) A1(M) A2(M) A3(M) A4(M) A5(M) B(M) A1(M) A2(M) A3(M) A4(M) A5(M)
>
>Simple main effects within group can then be assessed using appropriate
contrasts
>e.g. -1 +1 0 0 0 0 0 0 0 0 0 0
>...for A1-B in males
>
>Main effects of condition (across sexes) can be assessed using the above
contrasts concatenated across groups:
>e.g. -1 +1 0 0 0 0 -1 +1 0 0 0 0
>...for overall A1-B
>
>Group differences can be assessed using interactions of the above contrasts
>e.g. -1 +1 0 0 0 0 +1 -1 0 0 0 0
>...for areas where M's activate more than F's with condition A1 against B
>& +1 -1 0 0 0 0 -1 +1 0 0 0 0
>...for areas where F's activate more than M's with condition A1 against B
>
>Inference extends to the populations from which the samples were drawn,
>but only for these simple contrasts comparing pairs of conditions
>within group, and looking at differences of pairs across groups. ...and
>*only* in situations like this one where there are no replications of
>condition within subject. (One scan per condition per subject is the
>rule!)
>
>More general contrasts, such as:
> -1 +1/2 +1/2 0 0 0 0 0 0 0 0 0
>...are *invalid* for population inference, since they imply that A1 &
>A2 are to be treated as though they were the same condition (but
>fitting condition by replication interaction terms).
>
> ----------------
>
>In situations with replications of conditions within subject,
>especially in fMRI, the data must be boiled down to a single
>representative (adjusted) condition image per subject, and the
>resultant summary images assessed. A "kit" for doing such adjustments
>will be available shortly.
>
> ----------------
>
>Hope this helps,
>
>-andrew
>
>+- Dr Andrew Holmes [log in to unmask]
>| -___ __ __ Wellcome Department of Cognitive Neurology - |
>| ( _)( )( ) Functional Imaging Laboratory, Stats & |
>| ) _) )( )(__ 12 Queen Square, Systems |
>| (_) (__)(____) London. WC1N 3BG. England, UK |
>+---------------------------------------http://www.fil.ion.ucl.ac.uk/-+
>
--------------------------------------------------------------------------
Luca Finelli
--------------------------------------------------------------------------
Address: Institute of Pharmacology |Last night I lay in bed
University of Zuerich |looking up at the stars in the sky
Winterthurerstrasse, 190 |and I thought to myself,
CH-8057 ZUERICH |where the heck is the ceiling?
|
Phone: +41 (0)1 635 59 61 |(Dilbert's Words of Wisdom)
Fax: +41 (0)1 635 57 07
WWW: http://www.unizh.ch/phar/sleep
--------------------------------------------------------------------------
Have you seen our new BITS OF SLEEP CD-ROM? http://www.sleepcd.com
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|