Hi,
Setup:
Fast ER design with 8 conditions, 6 sessions, TR=1.5
I defined FIR as "Basic Functions" in 1-st level analysis. FIR length 18 and the
order parameter 12 (FIR length / TR length). Attached design matrix (FIR.jpg).
SPM estimated me a beta per each non-zero trial. In other words I got 96
regressors per sesssion (8 conditions x 12 repeatitions). Exactly the same
model with HFR definition doesn't suffer from this problem (attached HRF.jpg).
Interestingly, when I just extract percent signal change (without defining the
contrasts) they look as normal BOLD shape, though I think there is no way it's
a correct signal.
Did I make something wrong with FIR definition? I am not sure whether there is
any advantage for using FIR in such a design like mine. For example, this
comparison didn't answer me what I have to use:
http://www.fil.ion.ucl.ac.uk/spm/doc/papers/rnah_choice.pdf
Thanks,
Vadim
|