Dear Carsten,
What is the covariance 'Cy'? In 'spm_reml.m' I only see 'YY', 'C' and 'Cq' covariances.
I would be also happy to understand the ReML/pre-whitening procedures in SPM better. 2012 there was the study 'Does parametric fMRI analysis with SPM yield valid results? - An empirical study of 1484 rest datasets' showing very high (false) positive rates for FCP resting state scans with short TRs. I have recently worked with task and rest data from the enhanced NKI dataset, where some scans are at TR=0.645s, and I have also observed a lot of significant activation throughout the entire brain, both for FSL and SPM. FSL is believed to model auto-correlations more flexibly than SPM. However, the FSL 2001 study 'Temporal Autocorrelation in Univariate Linear Modeling of FMRI Data' referred only to TR=3s and TR=1.5.
Best regards,
Wiktor Olszowy
|