Hi - I entered this into FEAT and the ratio of min:max eigenvalues is
4e-2, ie not that small - hence your design is close to rank deficient but
not so close that I would expect you to have problems.
Cheers.
On Thu, 22 Apr 2004, mohanty wrote:
> I looked at the two matrices. The box on the lower right corner of the
> eigenvalues matrix is almost black. So, far in my testing, all of the
> eigenvalues for the design matrix are much bigger than zero (suggesting no
> linear dependency in the design matrix). I am not sure why I am getting the
> rank deficiency error. Also, I get the same rank deficiency errors as soon as
> I add any interaction terms to the matrix. Here's the matrix:
>
> ev1 ev2 ev3 ev4
> 1 -15 -33 -5
> 1 8 0 -9
> 1 12 26 32
> 1 14 25 17
> 1 14 25 -4
> 1 11 24 49
> 1 -17 1 -3
> 1 -19 0 -13
> 1 -17 -30 -16
> 1 14 2 -6
> 1 12 -23 -6
> 1 -16 -8 -15
> 1 15 1 17
> 1 14 8 -9
> 1 -17 -14 -13
> 1 -13 -18 7
> 1 -18 -21 -11
> 1 11 20 -10
> 1 16 13 7
>
> Thanks for you help.
>
> Appu
>
>
> >===== Original Message From "FSL - FMRIB's Software Library"
> <[log in to unmask]> =====
> >Hi - what you are doing sounds right. Is it possible that by some chance
> >the 3 predictor EVs _do_ actually add up to a constant? Have a look at the
> >"Covariance" - is the bottom-right square black?
> >
> >Cheers, Steve.
> >
> >
> >On Wed, 21 Apr 2004, Appu Mohanty wrote:
> >
> >> Hi,
> >>
> >> I am trying to do a regression analyses using the functional data as DV
> >> and three questionnaire measures as predictors. Since, the EV's are not
> >> demeaned at the higher level of analyses, I mean centered (EV value - mean
> >> of EV) the EV's beforehand. I then included an EV consisting of 1's
> >> (modeling group mean) and three mean centered EV's corresponding to the
> >> three predictors in the model. I get an error message stating "at least
> >> one EV is a linear combination of other..." I do not get this error if I
> >> include only the three EV's corresponding to the predictors and not the EV
> >> modeling the group mean. It appears that adding the EV modeling the group
> >> mean adds some linear dependency in the model. I am not sure why this is
> >> happening or how to proceed next. I would appreciate any suggestions.
> >>
> >> thanks
> >> Appu
> >>
> >
> > 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
>
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
|