Dear Lisa
Before considering the age question, if I understand correctly, this design matrix will be rank deficient - meaning that column 1 will be a linear combination of columns 2-4. You can't do that in a General Linear Model.
I'm happy to advise how to fix it - but please can I check, should your three groups be treated equally and separately? E.g. three groups of patients with three separate diseases? Or is there some implicit hierarchy to the groups - for example, is one group controls and two groups different kinds of patients?
Best
Peter
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Lisa Goudman
Sent: 21 November 2018 07:44
To: [log in to unmask]
Subject: [SPM] PEB covariates
Dear SPM experts,
We used DCM to explore differences between 3 groups of patients.
I'm running DCM PEB for EEG whereby the design matrix is constructed with the first column an average and columns 2-4 specific contrasts to observe differences between the three groups. Now, we added age as a covariate in the last column to "correct" for the age of each subject.
When we add age to the design matrix, then in the PEB-review parameters screen, there are less parameters with a high Pp for the first covariate (average). However, for the age covariate we get no Pp estimated parameters.
Is it correct to assume that the average and the specific contrasts are now corrected for the age of the person? And is it normal that we do not see estimated parameters for covariate age? We were thinking that maybe estimated parameters that are influenced by age would be showed? Or does that not make any sense?
Thank you very much.
Regards,
Lisa
|