Dear fMRI experts,
I have an event-related 6(cue type) x 2(target type) experimental design, in which each trial has a cue and followed by a target. The main aim is to examine the similarity and dissimilarity between cue-related multivoxel patterns. My pilot data, with a jitter ditribution between 3-7s between cue and target, and 3-7 s between target and next cue generated for each cue condition, has shown that within each scan session, the patterns are highly similar regardless of the cue and highly different across scan sessions. This made me worry that the signals between different events (ie., cue and target, and the next cue etc) within a scan session was not dissociated successfully.
Therefore, I have two questions:
1) (a)What would be an optimal jitter distribution distribution (min, max, and steps) for each condition, and (b) how many trials are needed for each condition in order to separate the signals successfully?
2) Looks like that even if one has an optimal jitter and sufficient trial number across the entire experiment, one can still end up not having enough power to dissociate the signals within each scan session if the experiment is segmented into multiple sessions. Thus one should think about the jitter distribution and number of trials at the level of session, instead of experiment. Is this right?
Any input is highly appreciated!
Many thanks in advance!
Jingwen
|