Mark,
> When single subjects are given multiple runs of the same trial, I understand the recommendation
> from FMRIB to be that the runs should not be concatenated directly unless there was extremely
> little motion. Instead, the recommendation is to bring the repeated trials of the individual to a
> second and third level analysis using FLAME.
>
> My question is: wouldn't doing so imply a mixed (random) effects treatment of the repeated trials,
> which should be a fixed effect?
Yes, it does imply that the trials are a random effect. Although I
wouldn't necessarily agree with you that it _should_ instead be a fixed
effect. The between trial variance component might be a worthwhile thing
to model.
> Would it not make sense to co-register the COPEs and VARCOPEs,
> average the COPE's and take the r.m.s. average of the VARCOPES (or some other transformation on
> the variances) to come up with summary stats on these single subjects to propagate to the next
> level analysis? This would have the added benefit of much reduced computation time.
Yes, if indeed you want to make it a fixed effect. Although, incidentally
doing a fixed effect analysis will be available as an alternative to
mixed effects in the next release of FEAT.
Cheers, Mark.
Mark Woolrich.
Oxford University Centre for Functional MRI of the Brain (FMRIB),
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
Tel: (+44)1865-222782 Homepage: http://www.fmrib.ox.ac.uk/~woolrich
|