Hi whole wide SPM world,
I am using SPM 12.7487 and calling spm_fmri_concatenate() to concatenate
over my 2 scanning sessions and 2 runs/session in order to extract a
single timeseries across all the 4 runs for connectivity analyses. For
one of my participants (works fine for the others), the first-level
model estimates fine when I don't concatenate the runs, but if I do
spm_fmri_concatenate() first, model estimation fails during
non-sphericity correction spm_reml (v7178) (AR(0.2)):
____________________ SPM error ____________________
K>> Index exceeds the number of array elements. Index must not exceed 0.
Error in spm_reml (line 62)
n = length(Q{1});
Error in spm_reml (line 213)
[V,qE,qP] = spm_reml(YY,X,Q(iQ),N,t,0,hP(1)*2);
Error in spm_est_non_sphericity (line 276)
[Vp,hp] = spm_reml(Cy(q,q),Xp,Qp);
Error in spm_spm (line 433)
[xVi, am] = spm_est_non_sphericity(SPM);
Error in proj1_41_designGLM_concatTS (line 418)
SPM = spm_spm(SPM);
______________________________________________________
triggered by Q being an empty cell
Tracing back, at spm_reml() l.213, when spm_reml() calls itself in the
block to "check V is positive semi-definite and parameters have been
identified", in the second iteration through xX.K
K>> min(eig(full(V))) = -0.1478
and iQ is an empty cell
Not sure what I'm looking at in this block nor what to look for/where
the problem lies.... My model looks correctly specified, as far as I can
tell, at least, but I guess wrong model design is the most likely
problem? I have attached the design matrix plots with no concatenation
and after concatenating + spm_fmri_concatenate() in case.
Many thanks -- any help/leads/remote explanations very much appreciated!
Gina
Does "v7178" literally mean the 7178th version?
|