Hi - if you design was close to rank deficient, I wouldn't bother trying
to interpret the results, no - I would just re-run with a non-deficient
model.
Cheers.
On Fri, 29 Oct 2004, Ongur, Dost,M.D. wrote:
> This is very helpful, thanks. In a design such as mine that is close to being
> rank deficient, do the PEs deviate in a predictable way (e.g. all become smaller
> in magnitude)?
> In other words, if I want to find out whether there is something interesting in
> the patterns that I saw with my PEs, is re-doing the design the only way?
> Thanks,
> Dost
>
> -----Original Message-----
> This is a little confusing - in a block design like this you should not
> generally model _all_ conditions - there will normally be one condition
> which you can consider "baseline" (I'm guessing CH in this case) which
> you
> should not model. This is because (at first level) the data gets
> demeaned.
> So you should probably only have EVs for A,B,C,D and use appropriate
> contrasts to ask whatever questions you want.
>
> > - All 5 PEs we specified in our design matrix (block design) are
> negative
> > in our ROI. Thinking that our ROI may have a lower blood flow level
> than
> > the brain overall as Joe Devlin mentioned, we ran avwstats on a couple
> > filtered_func_img and get a mean of about 9900. But the raw data in
> our
> > ROIs are typically around 11000-13000!
>
> I'm guessing that getting -ve PEs is a result of the design being close
> to
> rank deficient for the reasons above. Contrasts would still be well
> conditioned, but PEs on their own would not be. I suspect this is what's
> going on.
>
Stephen M. Smith DPhil
Associate Director, FMRIB and Analysis Research Coordinator
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|