Hi all,
I would like to get opinions from SPM and PPI experts on the following
analysis I am doing:
I have done an activation analysis for an event related design, using FIR
HRF at the single-subject level. I would like to add connectivity (PPI)
analysis to this design. I am not finding much literature on how to do
this.
Briefly, the process of creating a PPI vector involves the following steps:
1. Deconvolve the BOLD time series (first eigen variate of voxel time
courses from seed region) by the HRF
2. Multiply with psychological factor
3. Convolve with HRF
In the case of a FIR HRF, I am using time bin=TR, so each column of the HRF
is just a delta function with a delay.
So deconvolution(step 1) is advancing the BOLD time series by the
corresponding delay time, while convolution (step 3) is delaying the result
of multiplication by the same amount. Therefore, the PPI regressors can be
created just by multiplying the time course with the task related columns.
1) Is anything wrong with this reasoning?
2) What other option do I have to perform a PPI with FIR HRF?
3) Also, should I be doing a detrending for the PPI vector? If so, at what
stage?
Thank you for looking at this.
Uma
