Hi,

I have a query about the recommended approach for handling events with onsets that are prior to the data but close enough that their response has an influence.

For example, I have an ABAB..AB block design where I want to discard the first and last blocks - so that data at t=0 is the start of the second block. But the first few timepoints in the resulting data include the influence of the first block, which I'd like to account for.

The way I'm handing it at the moment is to generate a HRF timecourse for the initial block using 3dDeconvolve with '-nodata', manually cull the points corresponding to the first block, and pass the result to 3dDeconvolve using '-stim_file'.

As another example, I have an event-related design where the events follow an M-sequence of 124 events. Each run consisted of such a sequence, prepended by the final 8 events of the sequence. I would like to discard the data corresponding to the presentation of those initial 8 events, but account for their influence on the initial timepoints of the resulting data.

Is there a recommended approach for such situations? Including the full timeseries and censoring seems like an option, but also seems to be not censoring's intended purpose. There was a previous discussion about being able to pass negative '-stim_times' to 3dDeconvolve (http://afni.nimh.nih.gov/afni/community/board/read.php?1,77598,77598#msg-77598), which would seem to be the best solution if it is feasible?

Thanks,

Damien.