Hi Anda,
It shouldn't happen that just rescaling a covariate would change the rank
of the matrix, however in practice the ratio printed out by FEAT might
change _very slightly_ - if this is the case then it must have only just
failed the rankdef test before i.e. not be very rank deficient!
Cheers.
On Mon, 13 Sep 2004, Anda van Stegeren wrote:
> Just want to share some knowledge and question at the same time how this can be.
>
> In a large design, where subjects came for fMRI in 2 sessions, I want to enter a physiological measure as covariate.
>
> Created an extra EV and entered for each subject the demeaned value of the measure twice (one for each session).
> The design was rejected as "rank deficient".
>
> The graphical representation of the design matrix and parameter contrasts showed that the "red line" for some subjects/values was extremely to the right or left position, indicating a relatively large difference from 0 zero.
> there were indeed high values >10 and some >100 here.
>
> I decided to divide all values /100 and entered these values in a new design (demeaned/100 = still demeaned, I reasoned).
>
> Now it was accepted ! without rank deficiency messages.
>
> 1) If this is correct: then it can be of interest for all users, that these demeaned values for covariates should not exceed certain numbers.
>
> 2) I would like to know if there is any argument against this procedure.
>
> best regards,
> Anda
>
>
> Drs. Anda van Stegeren,
> University of Amsterdam
> Department of Clinical Psychology
> Roetersstraat 15
> 1018 WB Amsterdam
> The Netherlands
> tel. ++ 31 20 525 6799
> fax. ++31 20 639 1369
> E-mail: [log in to unmask]
>
--
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
|