Dear Michaël,
On the practical part of your question, I don't see any problem doing
what you want to achieve with SPM's first level model specification:
don't specify any conditions and select the FAST option for serial
correlation (FAST is only a name given to a particular set of variance
components - it's not an acronym and not many thoughts went into finding
an appropriate name).
See for example:
https://www.fil.ion.ucl.ac.uk/spm/doc/manual.pdf#subsection.38.2.1
Best regards,
Guillaume.
On 19/03/2019 01:27, Belloy Michaël wrote:
> Dear member of the SPM list,
>
>
>
> I would like to ask both a practical and theoretical question pertaining
> to autoregression in SPM.
>
>
> 1.
>
> I would like to implement F.A.S.T auto-regression within the multiple
> regression factorial design.
>
> Normally F.A.S.T is an option in fMRI model specification, in order to
> deal with serial auto-correlations.
>
> I would however like to generate resting state fMRI scans from which the
> motion parameters and serial auto-correlations are regressed.
>
> Does anyone have advice on how to achieve this?
>
>
> 2.
>
> I am trying to understand what is the fundamental issue that
> auto-correlation deals with.
>
> Specifically, does auto-regression deal only with inherent signal
> properties (e.g. prior history from spin effects) that can
> 'falsely' increase ROI-ROI correlation or affect sensory responses?
>
> Or, does it also affect longer self-predictive components of the signal.
>
> To be clear on the latter, there are hypothesis that infra-slow brain
> rhythms may affect sensory responses through biological phenomena.
>
> If I have a particularly hypothesis related to such phenomena, than
> auto-regression would not be desirable?
>
>
>
> Thank you,
>
>
> Michael
>
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
UCL Queen Square Institute of Neurology
London WC1N 3BG
|