Thanks a lot, Steve.
However, wouldn't using the lower 5th percentile of the entire timeseries, instead values "relative to each stimulus" (e.g., the 3-5 volumes prior each stimulus), result in inaccurate estimation of the %signal change, at least in some instances?
For instance, let's hypothesize that my voxel is activated by a stimulus, and deactivated by a subsequent stimulus. In this case my lower 5th percentile would likely fall during the voxel's deactivation. If I take that value as my "baseline", I would end up overestimating the magnitude of the activation, and underestimating that of the deactivation. Or am I missing something?
Thanks for clarifying this.
Marco
|