Low frequency and long time, epoch

Hello Mike,

When I watched your video, I cannot remember where it is. I still remember that you talked about epoch . When we try to analyze epoch with low frequency (eg. 4-7 Hz), you said that it requires a bit longer time. My question are :

  1. Why do we need such a longer time, eg. 2 seconds, for low frequency ?
  2. How about high frequency (eg, 20-30Hz), does it mean that it does not need a longer time per epoch ?
  3. What is actually the best duration for each epoch ?
  4. Does the length of time per epoch affect a result ?
  5. Is it common to adjust a epoch if a result does not fit ?
  6. Why do people still need a epoch, even though during the experiment there is no such epoch, say recording 30 minutes of brain wave of sleeping ?
  7. However, if my understanding is wrong and not in line with what you explained in the video, please correct me.

Thank you and sorry for the many questions.

Thanks a lot

Hi Ihshan. This is about edge effects. Any time you apply a temporal filter, there will be distortions or corrupted signals at the boundaries (and sharp edges). These edge effects can last a few cycles, so I generally recommend making sure there is ~3 cycles of the narrowband filter frequency to absorb the distortions. Generally, if you have 1 second of extra padding on either end of the epoch, you should be fine.
I talk about this in multiple courses, and have quite a bit about this in my ANTS book.

Thank you so much for the explanation. So that should be fine in case I focus on theta (4-7 hz) with 2s for each epoch, right ?