Dear Kristine,
At 16:25 30/09/98 +0100, you wrote:
| Our task is to compare a set of pathological patients with normals on a
| cognitive task. I have 63 images taken per participant, 9 scans per
| epoch, with the tasks administered as 1 2 1 2 1 2 1, where 1 = rest
| condition and 2 = cognitive task condition. I have been instructed not to
| average over the subjects in each group (for fear of losing information),
| and we have 2 groups, 20 subjects in total.
Presumably you're interested in assessing whether patients with the
pathological condition in general are different from normals in general,
rather than just testing whether these particular patients are different
from these particular normal controls at the time they were scanned.
The former requires a random effects analysis, which can be effected by
summarising the data within each subject into one scan for each of the two
conditions. In the context of inferring about population differences,
repeat observations on an individual are of lesser relevance, and no
relevant information is lost by averaging the data in this way.
The latter is a fixed effects analysis. For group comparisons it is very
hard to justify an approach which only infers differences between the
patients & normals at hand. Most discerning journals will now reject group
comparisons effected with such an analysis when the conclusions attempt to
generalise the results to the corresponding patient populations.
The random effects kit for SPM96 will get you started: See:
http://www.fil.ion.ucl.ac.uk/spm/spm96.html#RFX96
| Additionally, the subjects have been matched to control for several
| nuisance variables, and I do not know how to enter this information into
| SPM.
This is interesting. This is theoretically quite easy to do in the
multi-level framework: You just compute (adjusted) difference images for
each subject and then pair the subjects in a simple model by including a
block effect for the pair of images. Unfortunately, due to limitations in
SPM96 on handling derived images, this won't be possible until SPM98. (The
problem is that SPM96 stats uses a proportion of the global mean (after
scaling) as threshold to identify intra-cerebral voxels for analysis. With
derived images, such as difference images with negative values, this isn't
appropriate.)
Hope this helps,
-andrew
+- Dr Andrew Holmes [log in to unmask]
| ___ __ ___ Robertson Centre for Biostatistics |
| ( ,) / _)( ,) Boyd Orr Building, |
| ) \( (_ ) ,\ University Avenue, |
| (_)\_)\__)(___/ Glasgow. G12 8QQ Scotland, UK. |
+----------------------------------------- http://www.rcb.gla.ac.uk/ -+
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|