This routine creates a schedule with a single non-null event type for a run with 100 time points with TR = 2 sec. The trial type will have the label 'evt1', will last 2 sec, and will be repeated 30 times. The 3 best schedules will be kept. Set the PSD window to 20 sec (which should be enough to capture the response to a 2 sec event). Note that the dPSD is implicitly set to the TR because there are only two parameters that follow --psdwin.
optseq2 --ntp 100 --tr 2 --psdwin 0 20 --ev evt1 2 30 --nkeep 3 --o ex1 --nsearch 10000
There will be five files created by this process. The three schedules will be stored in ex1-001.par, ex1-002.par, and ex1-003.par. Each file will have 4 columns corresponding to: (1) stimulus onset time, (2) event type numeric id, (3) event duration, and (4) event type label. ext1.log is a file that contains information about the progress of the optimization. It is created when the process starts, and the process adds to it after it has completed each 10% of the search or when one of the best schedules changes. ex1.sum is a summary file with information about how optseq2 was run, statistics about the search, and statistics about each of the kept schedules.
Look at one of the event schedule files (eg, ex1-001.par). Notice that while the duration of event type 1 is constant at 2 seconds, the duration of the Null event can change (but will always be a multiple of the dPSD).
Look in the summary file (ex1.sum). What was the efficiency for the best schedule? Re-run with 60 and 90 repetitions. How does the efficiency change? How high can the number of repetitions go before you encounter a time violation?
This routine creates a schedule with three non-null event types for a run with 120 time points and TR = 2.5 sec. Event type 1 will have a duration of 2.5 sec with 20 repetitions. Event type 2 will have a duration of 5.0 sec with 10 repetitions, and event type 3 will have a duration of 7.5 sec with 15 repetitions. Set the PSD window to 25 sec to assure that the response to event type 3 can be captured; explicitly set the temporal resolution (ie, the dPSD) to one TR.
optseq2 --ntp 120 --tr 2.5 --psdwin 0 25 2.5 --ev evt1 2.5 20 --ev evt2 5.0 10 --ev evt3 7.5 15 --nkeep 3 --o ex2 --nsearch 1000
Look at one of the event schedule files (eg, ex2-001.par) to verify that event onsets always occur at integer multiples of the dPSD (2.5). What was the efficiency of the best schedule?
Change the dPSD to half a TR (ie, --psd 0 25 1.25). How did this change the efficiency? Verify that event onsets always occur at integer multiples of the new dPSD, and that the Null durations are always an integer multiples of the new dPSD.
Create a DOF violation by changing the dPSD to .0625.