Hi Mark, good to hear from you.
I assume that you only have one number for the behavioural measure per
subject (and not a time series of values that could be fed into the
first-level analyses). In which case the second level analysis should
probably be like the example on the web
http://www.fmrib.ox.ac.uk/fsl/feat5/detail.html#SingleGroupAveragewithAdditionalCovariate
(the "Single-Group Average with Additional Covariate" section)
Cheers, Steve.
> Seems a simple question, but I haven't figured the answer. My goal is to
> do a parametric group analysis where each subject was given the same
> task, but behaved differently - we wish to find the locations whose
> signal co-varies with our behavior measure - really, a simple
> correlation question. Can you confirm that the following is the correct
> design (or suggest the correct one)?
>
> First level: EV1 is just the Task-Rest
> Second Level: EV1 is the de-meaned per-subject value of our behavior
> measure
>
> Questions:
> 1) As correlation does not reqiure a de-meaned EV, why should this
> study? 2) Is there a difference between this model, and in bringing in
> our behavioral measure at the first level (or are the COPEs simply
> scaled)?
>
> Thanks, as always.
|