I am using noise covariance matrix that was calculated from the resting state data in order to compute the inverse solution and is not filtered at all. However, my data is currently filtered between 1 and 30 Hz. My question is that using ulfitered noise covariance matrix for the filtered data will be cause of any potential problem or not ?
Thanks in advance
- MNE version: e.g. 0.23.0
- operating system: e.g. Linux Centos 7 with Python 3.9.1
Hello, you should preprocess your “noise” as closely as possible to your “real” data to ensure it contains actually meaningful information for source estimation.
Whether resting state data is a good source of “noise” is another important question to answer. Generally, I would say “no”, depending on your experimental paradigm and planned analysis. If you want to calculate the inverse solution for resting-state data, use an ad-hoc covariance matrix.