[Eeglablist] Recovering single component's channel data manually

[Robert Gougelet]

I realized that I was confusing a component activation with its projection,
and it should be

comp_chan_data = EEG.icawinv(:,comp_i)*EEG.icaweights(comp_i,:)*EEG.data

Rob

On Wed, Aug 7, 2019 at 1:10 PM Robert Gougelet <rgougelet at gmail.com> wrote:

> I just wanted to verify the following:
>
> EEG.icaact = EEG.icaweights*EEG.data;
> EEG.data = EEG.icawinv*EEG.icaact;
>
> When applying an ICA decomposition to a dataset in earlier preprocessing,
> I want to see the channel-level activity of a single component. Is this the
> correct approach?:
>
> comp_chan_data = EEG.icaweights(comp_i,:)'.*EEG.data;
> EEG.icaact(comp_i,:) = sum(comp_chan_data,1);
>
> Where comp_chan_data is a nbchan x pnts or nbchan x trials x pnts matrix?
> So, if I wanted to look at the contribution of a single component at
> channel 15 I'd simply use comp_chan_data(15,:) or comp_chan_data(15,:,:)?
>
> Is polarity adequately "undistributed" from the ICA decomposition back to
> the channels with these transformations? Thanks.
>
> Rob
>