[Eeglablist] Number of PCs to use in the FMRIB tools

[Makoto Miyakoshi]

Dear Mariam,

Since PCA is agnostic about the nature of data, only you can determine the
threshold.
And it should vary from subject to subject.
If your priority is data quality, then you should check it from subject to
subject, varying number of PCs from 1 to 5 etc. If your priority is speed,
you can determine it with some criterion and need to be satisfied.

Note also that when you see PC loadings, it exponentially reduces as the PC
number goes up. In other words, the difference between 1 PC and 2 PCs and
that between 2 PCs and 3 PCs could be very large.

Makoto
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20170510/f6f78162/attachment.html>