Dear community,

I was looking for some advice on the appropriate parameters to use for the BLOCK model for a deconvolution.

Based on what I've read thus far, it seems BLOCK is appropriate for my situation, as compared to GAM, for example, though it is an event-related design.
The entirety of the stimulus is 4 seconds long, but it's internally complex. By that I mean that the relevant activation isn't likely to start at onset, but rather 2 second in. Participants are essentially answering a question that is posed at stim onset, the question lasts 2 seconds, then they are presented with two options (2AFC), 1 second each, then they have 3sec+jitter to register their response via response box.
Perhaps complicating matters is the fact that I'm interested in sub-cortical structures which have different HRFs/latencies than, for example, the neocortex.

First, is BLOCK appropriate in this case or is that a mistake?
Second, what value I should use for d? (I assume d is in seconds, btw, and not TR, right?)
Relatedly, what do you think the best times I should use as my stim-times?

Thanks for the advice or info on any similar experiments you've analyzed,
Marc