Dear SPM list
Assuming an experiment has two conditions, A and B, and three
sessions. The conditions depend on the subject's responses (e.g.
correct, incorrect), i.e. it can happen to have no event in a certain
condition in a session.
Subject 1 has events for conditions A and B in all three sessions.
Subject 2 has events for condition A in all three sessions, but for
condition B there was no event in the third session.
In the design matrix I exclude "empty regressors", so the design
matrix would look like this:
Subject 1: A B A B A B
Subject 2: A B A B A
(where A and B are subject-specifc vectors of event times)
Accordingly the t-contrasts (A vs B) would look like so:
Subject 1: 1 -1 1 -1 1 -1
Subject 2: 2 -3 2 -3 2
With these contrasts the con_NNNN images are somehow scaled
differently. Can I nevertheless use them for further analyses (e.g. a
one-sample t-test) or will there be an up-weighting of Subject 2?
Likewise the contrast 'B vs 0' (to feed into a flexible factorial)
would be as follows:
Subject 1: 0 1 0 1 0 1
Subject 2: 0 1 0 1 0
Again scaling seems to be different: the con_NNNN image of Subject 1
is the sum of 3 betas, where the one of Subject 2 consists only of 2
Do I have to correct for the fact that con_NNNN images consist of
different numbers of beta values?
For example by using scaled contrasts:
Subject 1: [0 1 0 1 0 1] / 3
Subject 2: [0 1 0 1 0] / 2
Or is this already automatically handled by SPM somehow?
This post is related to this one
where I could not find a conclusive answer to the question.
Thank you for any help
Christoph Hofstetter, MSc in Biology
Laboratory for Neurology and Imaging of Cognition
Dept of Neurosciences
University Medical Center (CMU)
1 Michel-Servet - 1211 GENEVA - CH