SME before or after ICA for Bad Channel Identification

[samuelsze333]

Description

I would like to consult about using SME in ERPLAB to identify bad channels. I have gone through "7.6: Exercise - Interpolating Bad Channels" of "Applied Event-Related Potential Analysis" written by Prof. Luck. In the exercise, the sample dataset has been preprocessed (rereferenced and high-pass filtered). And then the instruction guides us through the SME computation.

My first question is: When should we compute SME and identify bad channels, say, if we follow the example pipeline in "14: Appendix 3: Example Processing Pipeline"? Should we apply it after ICA and epoching? Or should we epoch data and then compute SME before ICA?

Second, we have six bins. In only one bin, an electrode is 2 SD away from the mean for most 100-ms time windows. Should we interpolate this electrode given that we want to analyze it?

Third, if an electrode of interest is 2 SD away from the mean for just one 100-ms time window of interest, is it true that it is not worth interpolating this electrode?

Fourth, are there any thresholds regarding the proportion of (1) time window and (2) bin where an electrode is 2 SD away from the mean for that electrode to qualify as a bad electrode for interpolation?

Thank you so much for your help in advance!!!