I'm analyzing one patient's MEG, and I want to know the source for a specific frequency band.
The epoch of my trial is [-4s, 4s] and the stimulation started from 0s. I just want to know the activity of [0s, 4s].
So after source estimation, I've tried these two orders of analysis:
first applied band-pass filter of source, then extract [0s, 4s] from the filtered source;
first extract [0s, 4s] from the source, then applied band-pass filter to the extracted source.
I found that the results have big differences. I'm wondering which processing order is correct.
Use Option 1. You should apply the filter on the longest signal, this is to avoid transient artifacts, which are likely the differences you are observing.
The duration of the transient artifact depends on the filter design. The sharper the filter, more tabs are needed, which leads to longer transient artifacts. Check this tutorial to have deeper insights on the use of digital filters: https://neuroimage.usc.edu/brainstorm/Tutorials/ArtifactsFilter