The recently added DM function in AFNI is nice, but some additions will be very useful. Grinband et al. (2008) show that DM may capture RT effects better than AM. However, by DM, they mean an HRF convolved with a boxcar function (as long as the RT). This modulates both the duration and the amplitude. The amplitude is modulated to a lesser extent than AM (RT twice as high will give a peak height a little less than twice as high; see Fig. 6a in the paper). The dmBLOCK function modulates only the duration, while keeping the amplitude fixed. Using AM and DM at the same time also does not replicate the Grinband method, because AM will do a "full" modulation of the amplitude (RT twice as high = peak twice as high).
What is needed is a method of convolving different length blocks with some standard HRFs. This will be similar to the BLOCK function already there, except the length of the block will vary according to the parameter after : in the file.
On a related note, can convolution be allowed for multiple HRFs? Maybe with a regular GAM, an incomplete GAM, and an HRF with an undershoot (e.g., the SPM for FSL HRFs).
Also, it seems that currently the only option for a single standard HRF is GAM. SPMG has the canonical and the temporal derivative, and SPMG3 has one more. Could we have an option for just the SPM canonical function without the derivative (although that can be implemented with EXPR by the user)?
Thanks!
Ref: Detection of time-varying signals in event-related fMRI designs
NeuroImage, Volume 43, Issue 3, 15 November 2008, Pages 509-520
Jack Grinband, Tor D. Wager, Martin Lindquist, Vincent P. Ferrera, Joy Hirsch