Dear all,
As for my previous post, Dr Henson has been so kind to provide me with some advice in:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1110&L=SPM&F=&S=&X=5AE4E7267D9D7592CE&Y=mocheck%40sina.com&P=209808
However, I would like to make sure that I understood correctly.
1) As Dr Henson mentioned, if I am interested in response vs baseline, I really don't have to care so much about reducing SOA length. In fact, I have to include null trials which I suppose, will result in longer SOAs?
Also, can I understand it this way that within a fixed period of experimental time shorter SOAs lead to more trials than longer ones, which in turn yields more scans? In other words, if I include enough trials per event type, longer SOAs do not really matter in terms of statistical efficiency?
2) Dr Henson also mentioned that one could model nonlinearities with Volterra kernels and in his writing, he also mentioned that the overlap between successive events can be modelled via HRF. Actually I am quite curious about these two possible modeling ways. My questions are:
2.1) Will modeling overlap and nonlinearities yield better results?
2.2) Has anyone done these before? If so, are there any articles or tutorials about how to model these two effects?
3)Do I still need to insert jittered period of 2.75-3.25s between 2 consecutive trials. To what I know, if I am interested in comparison between different event types, these jittered periods are required. If so, is the range centered at 1/2TR( e.g. [1/4 TR, 3/4TR]) acceptable?
Any comments and advice will be greatly appreciated.
Many thanks and Best regards
Ce
