SSP _pca

Hello Francois,
I have a question when i use SSP:generic.SSP-PCA
From the picture, we can see the SSP1(27%) but i think it isn't nosiy. Does that mean the SSP:generic is not suitable for my EEG database? Thank you!

SSP is not always a good choice for EEG preprocessing.
If you don't get anything that makes sense with SSP, try with ICA:

is it common practice to select SSP Components manually? Are there any alternative automatic methods?

Yes, that's common practice, just like ICA components. Of course it is not satisfactory because it is user-dependent but for most sources of nuisance, removing the first one or two components guided by their respective topography and signal traces is a pragmatic and justified approach.

1 Like

I see. Thank you.
Does SSP in brainstorm only illustrate first 20 components, because I am looking into process_ssp2.m file where SSP happens, and it appears that U in [U,S,V]=svd(F) is 340x1 matrix (meaning we have 340 components to choose from).?


That's right but the first few components are those that capture typically 99% of the data variance.

1 Like

Thank you Dr. @Sylvain.
Are the shown components sorted or just within the order of SVD computation. ?
If I recall correctly, in ICA the negative positive or value of Eigenvalues were not the indicator of their importance .. but not sure about SSP


They are ranked in decreasing order of variance explained (shown as %tage values next to every component)

1 Like