I am trying to run the fastICA on my data. However, I am facing some issues.
- To what I got, usually the fastICA produces 5 outputs (i.e., the pre-processed data matrix X, the pre-whitening matrix K, the estimated un-mixing matrix W, the estimated mixing matrix A, and the estimated source matrix S). I am not sure which matrix is generally used (and particularly in this script: [https://nilearn.github.io/auto_examples/05_advanced/plot_ica_resting_state.html#sphx-glr-download-auto-examples-05-advanced-plot-ica-resting-state-py] ) to extract the components.
- Should I then apply the threshold just to the matrix of the components? And is there an optimal threshold that I should choose?
Hope you will help me to shed light on these issues.