Dear Stefan
The AR(0.2) reflects that the AR(1) process in SPM is actually a linearised version of an AR(1) process namely one with a fixed parameter 0.2. What the modelling in SPM is then about is finding out how much of the AR(1) with fixed parameters (and how much of a white noise process is needed).
So lets call it a feature :-)
Best
Torben
> Den 28. jul. 2016 kl. 16.16 skrev Greulich, Stefan <[log in to unmask]>:
>
> Dear SPMers,
>
> I am having problems with the correction for autocorrelation. While digging through the code of SPM12, I found that, even if AR(1) is selected in the GUI, the switch loop in spm_fmri_spm_ui.m line 237 goes into the otherwise case and prepares an AR(0.2) model.
>
> I would like to post, if it makes a difference. Manually setting matlabbatch{3}.spm.stats.fmri_spec.cvi = 1 should trigger line 231 in spm_fmri_spm_ui.m and use an AR(1) model. However, in this case the model estimation fails with:
>
> Failed 'Model estimation'
> Undefined function or variable "dh".
> In file "/Users/Documents/MATLAB/spm12/spm_reml.m" (v5223), function "spm_reml" at line 106.
> In file "/Users/Documents/MATLAB/spm12/spm_reml.m" (v5223), function "spm_reml" at line 181.
> In file "/Users/Documents/MATLAB/spm12/spm_est_non_sphericity.m" (v6015), function "spm_est_non_sphericity" at line 264.
> In file "/Users/Documents/MATLAB/spm12/spm_spm.m" (v6678), function "spm_spm" at line 421.
> In file "/Users/Documents/MATLAB/spm12/config/spm_run_fmri_est.m" (v5809), function "spm_run_fmri_est" at line 33.
>
> Is that a bug or a feature?
>
> Cheers,
>
> Stefan Greulich
>
>
> --
> Stefan Greulich
>
> Deutsches Primatenzentrum GmbH
> Kellnerweg 4
> 37077 Göttingen
>
> Tel: +49 551 3851-425
> LinkedIn: https://www.linkedin.com/in/stefangreulich
|