Hi Bruno
for something quite different, I had this question too ..
- what I did is to edit spm_defaults and comment some parts of spm_spm
(toward the end) to generate residuals - so I compute my glm with no
conditions, just motion and AR(1) and filter -- that way I get
time-series with no 1st order correlation, filtered, and motion
corrected -- from there you can do other stuffs :-)
Cyril
> Dear list,
>
> this question is closely related to a recent thread I started on
> computational issues in permutation-based tests of the significance of
> parameters derived from a 1st level GLM.
>
> I am analyzing an event-related design with many different
> stimulus-specific conditions. The parameters of interest to me are
> derived from the condition-specific betas of the 1st level GLM. The
> test of my hypotheses relies on permutation null distributions based
> on the exchangeability of my stimulus-condition labels (i.e., the null
> relies on condition-specific betas estimated by shuffling the stimulus
> labels assigned to each of my trials).
>
> One of the issues with this approach I have discussed offline with
> Donald McLaren (thank you!) concerns the estimation of the
> non-sphericity parameters and of the whitening matrix: should they be
> re-estimated independently for each permutation sample, or not? The
> reasonable answer for the moment appears to be that the whitening
> matrix should be constant across the permutation samples, i.e., should
> be estimated once from the un-permuted 1st level model. However, and
> here is my question, I wonder whether whitening is actually necessary
> for my permutation analyses: in the absence of whitening, serial
> correlations in the fMRI time series will express themselves in both
> the "plugin" parameter estimates (un-permuted stimulus labels), and in
> the permutation parameter estimates used to build the null. For this
> reason, I am thinking that whitening might actually not be necessary
> for my specific case.
>
> On a related question, I am wondering the extent to which whitening
> itself might affect the fit of HRFs to event-related fMRI time series.
>
> Thank you for your opinion/feedback, and sorry if my question goes
> beyond the usual computational ones.
>
> Best,
>
> Bruno
>
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Bruno L. Giordano, PhD
> Institute of Neuroscience and Psychology
> 58 Hillhead Street, University of Glasgow
> Glasgow, G12 8QB, Scotland
> T +44 (0) 141 330 5484
> Www: http://www.brunolgiordano.net
>
>
>
--
Dr Cyril Pernet,
Academic Fellow
Brain Research Imaging Center
http://www.bric.ed.ac.uk/
Division of Clinical Neurosciences
University of Edinburgh
Western General Hospital
Crewe Road
Edinburgh
EH4 2XU
Scotland, UK
[log in to unmask]
tel: +44(0)1315373661
http://www.sbirc.ed.ac.uk/LCL/
http://www.sbirc.ed.ac.uk/cyril
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
|