Dear all,
I am doing a MVPA analysis on a sparse event design with motor actions.
I want to analyse the data for different time points with the FIR model
and correct for reaction times, by including them as a condition of non
interest, modelled with the HRF function with parametric modulation. I
tried with a parametric modulated FIR function, but this makes my
contrast non estimable, as subject report and RT have the same onsets.
My idea is now to construct the regressor from the condition myself and
include it as an regressor of non interest. I found that SPM calls the
function spm_Volterra, spm_get_bf and spm_get_ons However, the require
an existing SPM structure as input and model again all conditions with
the selected basis function.
My questions are as follows:
Has anyone tried anything similar and would share their code?
Is just calling convolve on the modulated onset function sufficient? SPM
upsamples the Design matrix to the microtime resolution and then
downsamples it again to TR, however, my task design is synchronized to
the scanner onset, therefore this should not make a large difference for
me, right?
Many thanks and stay healthy everyone!
Stefan
--
Stefan Greulich
Doktorand
Abteilung Neurobiologie
Deutsches Primatenzentrum GmbH
Kellnerweg 4
37077 Göttingen
Tel: +49 551 3851-488
LinkedIn: https://www.linkedin.com/in/stefangreulich
|