> From [log in to unmask] Fri Jul 21 13:36:14 2000
> Date: Fri, 21 Jul 2000 14:35:38 +0200 (MET DST)
> Subject: covariate-timing
> From: Peter Erhard <[log in to unmask]>
Dear Peter,
> we just stumbeled over a question that might have been discussed
> earlier. Our fMRI-paradigm is slightly asynchronous to the image
> acquisition. From the performance of execution of the paradigm we
> derive a covariate. As during the images at the beginning and end of
> the block the execution starts or ends, we do not know what to choose
> as a covariate for these images (for all the other images, we use
> either the average of the performance data acquired during the image or
> zero (during rest). In this context we also would like to know,
> whether the covariate data is convolved with the hemodynamic delay
> function, and if so, how ?
If the covariate is presumed to have a neuronal correlate then it
should be convolved with the HRF (because this is how that effect will
be expressed in the data). One way to model this is to treat your
block design as an event-related design, where each block is a train of
trials or events. Using the parametric option simply modulate each
trial with the perfomance measure for that trial. This approach
eschews any problem with asynchrony between acquisition and trials and
automatically ensures approoriate convolution with the HRF in
'microtime' (i.e. time-bins of TR/16).
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|