Hello,
I understand that PLS does not make an assumption about the shape of the HRF (McIntosh et al. 2004). I also understand that the preference is usually to include a large time window to capture the whole HRF (e.g., 16 secs).
Hypothetically, let's say there is a design with 4 conditions and a control task. Each of these event lasts 4 secs and the ISI is 0.5 to 2.5 secs. The control task is interspaced between each experimental condition.
The default might be to go for a time window of around 15 secs (depending on TR duration), but in this case, that would include a portion of another condition. Because of this, would a shorter time window make more sense? E.g., 9-12 secs? The scenario is puzzling in part because it might be that 3 of the 4 experimental conditions are linked at a certain level. Since there's an equal number of trials per condition, that means the second trial might or might not be related to the first, and the proportion of the likelihood changes based on whether it's the distinct condition or the three other conditions.
Any thoughts most welcome!
A.
Hi Annik - It might be worth trying a few window sizes to see if you get the sort of "spill over" you are concerned about. So long as the differences in the first four or five TRs are robust, extending it furhter probably won't make much difference.
Baycrest is an academic health sciences centre fully affiliated with the University of Toronto
Privacy Statement - Disclaimer - © 1989-2024 BAYCREST HEALTH SCIENCE. ALL RIGHTS RESERVED