To whom it may concern,
I have two experimental phases in my study and I would like to calculate separate noise and data covariance for each of the two. First I would like to know whether this is a recommended approach, or should I merge the covariance across the two phases. Second, I cannot get it to work, when I calculate the covariance of one phase for a subject, the file is put under common file and with the name “Noise/data covariance”. When I try to calculate the covariance from the other phase, while selecting keeping both matrices, it outputs an “unknow error”. I suspect it was caused by the name/comment of the covariance matrices being the same, so I tried to rename the first covariance matrix and to my surprise, I cannot do that either.
Thanks,
HY
Each folder can only contain one noise covariance matrix and one data covariance matrix.
Note that the data convariance is only useful if you want to use the “beamformer” option, which is still not fully documented (to be used only if you really know what you are doing).
Renaming noise or data covariance matrices is not possible, and I’m not sure what effect you would expect from this.
If you want to use separate noise covariance matrices for different recordings, then you have to configure your subject differently: double-click on your subject and select “Default channel file: NO, use one channel file per acquisition run”, which is always the recommended option). This will allow you to store/move your separate conditions in separate folders, with separate noise covariance, head model and inverse model.
Note that these matrices are supposed to represent “sensor noise”. Does it really make sense for you to use different noise covariance matrices for the two parts of your experiment?
All the guidelines we have to give on this topic are available in the noise covariance tutorial:
https://neuroimage.usc.edu/brainstorm/Tutorials/NoiseCovariance#Variations_on_how_to_estimate_sample_noise_covariance