Explained variance and interactive PCA

I'm working with an EEG signal (129 channels) and I have to process it with the eeglab toolbox. I have to do a dimensionality reduction using the interactive PCA; in this way I should find the number of component to use with the ICA.

I need to have at least the 95% of explained variance.

Using the GUI I have done a decomposition using the ICA typing: 'g','tanh','interactivePCA','on'.

Then I have to choose the components with the maximum and the minimum variance displayed.

How should I do this choice taking in account the 95% of the explained variance that I want to have? Is the process that I have done until here right?

Answer 1

[Prasanna]

Hi Giulia,

It is my understanding that you're working with a 129-channel EEG signal in EEGLAB and using ICA with 'interactivePCA','on' to reduce dimensionality. Your goal is to select enough principal components to keep at least 95% of the signal's variance before running ICA.

When 'interactivePCA','on' is used, EEGLAB shows a plot of how much variance is explained by each principal component. To meet your goal, choose the number of components that together explain at least 95% of the total variance. This is done by finding the point on the plot where the cumulative variance reaches or exceeds 95%. You can just select the top N components that cover the required variance.

In summary, you can focus on how many components together explain 95% of the data. The PCA plot during the process will guide this decision. For more details, refer EEGLAB’s ICA for artifact removal documentation: https://eeglab.org/tutorials/06_RejectArtifacts/RunICA.html

Hope this helps!