Hi Chris,
> I have a problem with estimation in SPM5, which only occurs for some of my
> data and not others. When I estimate, the design matrix changes from one
> that is uniquely specified to one that isn't (see attached images). Hence
> I cannot view my results. Does anybody know what might cause this? Any
> information would be greatly appreciated.
what are you trying to analyse?
Are you interested in the differences between cert_1 and cert_2 or are you interest in the difference between expectation F3_1 & F3_2?
Since the values for F3_1 and F3_2 seem to be identical within each subject, this design is indeed not really estimable. Currently, you are modelling two hypothesis at the same time, i.e. that there is an overall difference between cert_1 and cert_2 and that there is a difference between F3_1 and F3_2 in relation to cert_1 and cert_2. That can't be estimated since you can not differentiated whether the observed changes are caused by an overall change from cert_1 to cert_2 or by a change from F3_1 to F3_2. That would be different, e.g., if F3_1 and F3_2 would also vary within a subject.
As far as I understood your design, the most reasonable thing to do would be to collapse the two covariates into one covariate (as SPM in fact did).
Good luck,
Karsten
--
-------------------------------------------------------------
Karsten Specht, PhD
Department of Biological and Medical Psychology
& National Competence Centre for functional MRI
University of Bergen
Jonas Lies vei 91
5009 Bergen
Norway
Tel.: +47-555-86279
Fax: +47-555-89872
[log in to unmask]
http://fmri.uib.no/
|