I am working my way through GLM notations and design
matrices, and I have a question:
I can design a oneway within subjects anova by either
selecting that option from BASIC menu, or doing
PET: mult subj: cond and covariates with 0 covariates. The design matrices
appear to be the same (2 subjects, 4 condition, see
attach SPM.mat files).
If I want to account for non-sphericity with replications
over subjects, though, the estimation process fails if I have
used the PET version, but succeeds with the BASIC version.
Are these design matrices truly equivalent, or is this some kind
of bug or ignorant-user feature?
Kerrie Tainter
PETIC
Biomedical Research Foundation
LSUHSC-Shreveport
[log in to unmask]
the error message I get is :
...
In spm_reml at 115
In spm_spm at 810
ReML Iteration : 31 ...NaN
Warning: Matrix is singular to working precision.
>In spm_reml at 77
In spm_spm at 810
Warning: Matrix is close to singular or badly scaled.
Results may be inaccurate. RCOND = -1.000000e+000.
>In spm_reml at 82
In spm_spm at 810
Warning: Matrix is singular, close to singular or badly scaled.
Results may be inaccurate. RCOND = NaN.
>In spm_reml at 115
In spm_spm at 810
ReML Iteration : 32 ...NaN
SPM2: spm_spm (v2.66) 16:58:58 - 01/08/2006
=======================================================================
Initialising parameters : ...computing???
Index exceeds matrix dimensions.
Error in ==> spdiags at 114
a((len(k)+1):len(k+1),:) = [i i+d(k) B(i+(m>=n)*d(k),k)];
Error in ==> spm_spm at 419
s = spdiags(1./sqrt(diag(s)),0,nScan,nScan);
Error in ==> spm_spm at 827
SPM = spm_spm(SPM);
??? Error while evaluating uicontrol Callback.
|