Dear Hongchuan,
Thanks for your detail message.
I have 10 subjects, and half of them are scanned "A B", the rest
are scanned "B A". I performed random effect analysis (RFX) on the whole
group, the "A B" group and the "B A" group. The results are totally
different from these three group, so I want to add a covariant for the
order effect in the whole group. For conjunction analysis, I want to
compare the difference of two scanner, but the main purpose is to
include them as a whole group to increase the number of subjects.
Regards.
Ze
-----Original Message-----
From: Hongchuan Zhang [mailto:[log in to unmask]]
Sent: Thursday, January 06, 2005 9:37 PM
To: Ze Wang
Subject: Re: [SPM] how to perform conjuction analysis and consider
scanning order?
Dear Ze,
If your subjects fewer than 10, you should select fixed-effect model
to include all the scans you have, then simply select contrasts you are
interested in by using ctrl key. However, this will limit your inference
for difference between subjects are not taken into consideration. If you
have more than 10 subjects, you should use random effects model. To
perform conjunction analysis (if I am right, you want to know the
conjunction effect of the two scanners), you should choose multiple
regression in Basic model, then enter contransts of single subjects by
defining covariates such as [1 1 ... 0 0 ...] and [0 0 ... 1 1 ...] to
represent contrasts obtained from the two scanners, then you could
define contrasts such as [1 0] and [0 1], and perform conjunction
analysis as above. There are two questions remained: (1) you should use
a model without constant term; (2) you should use a global null
hypothesis. Below you wiil find some links of help:
http://www.sph.umich.edu/~nichols/Conj/
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind9911&L=spm&P=R8437&I=-1&m
=2155
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0406&L=spm&P=R20455&I=-1
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0406&L=spm&P=R8373&I=-1
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0104&L=spm&P=R8273&D=0&I=
-1&m=2155
However, the solution above are based on SPM99, I am not sure it is also
fit with SPM2.
>Dear SPMers,
> I have two datasets scanned in different scanner using the same
>paradigm with different condition order. How to perform a conjunction
>analysis in SPM2 and how to define a covariate for the order of
>conditions?
>
>Thank you!
>
>Ze
>
Hongchuan Zhang
[log in to unmask]
2005-01-07
|