Hi,
In modeling a project, I've recently focused on ensuring accuracy in my onset timing and slice-time correction. I've come across a paradoxical result and I would greatly appreciated any help. I apologize in advance for the long post.
In SPM designs, a common goal is to set regressor onsets to the middle of the TR. SPM should do this automatically, based on settings of the microtime resolution and microtime onset (SPM.xBF.T0).
However, testing this has let me to find that adjusting the microtime onset (T0) shifts regressor timecourses in the *opposite* temporal direction than expected.
In comparisons at the subject level, models with T0=1 have regressors that are shifted later in time than models with T0=T/2, which are then later than models with T0=T. (Actual values: TR=2 s, num slices=38, T=14, varying T0 from 1 to 14.)
In a further test, I adjusted onsets manually. Manually adding 1 s (here, TR/2) to onsets in a model with T0=1 should be equivalent to unchanged onsets and T0=T/2. This is not what I see: manually adding 1 s at T0=1 does not match T0=T/2, instead they match a T0=1 model if you imagine the regressor shifted 1 s later.
I have tried to follow the SPM8 manual and message board posting advice (e.g. https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=spm;3b158be4.02 and https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0804&L=spm&D=0&P=198620) which makes the above seem paradoxical. Thus, the above result make me worried that there is an error of some kind in SPM itself. Alternatively, it's entirely possible that I have misunderstood the issue!
Thank you for any help!
Best regards,
Elliott
Learning Lab
Columbia University
312 Schermerhorn Hall
[log in to unmask]
|