Introduction on Functional Principal Component Analysis - Part 1

I recently gave an 8-hour short course on functional data analysis. These videos can be viewed as the updated version of my previous videos. This is the second session.

[longhaili]

wonderful videos! Thank you for making them available and public.

[markcarpenter]

Since the weight functions are orthogonal and the first weight function is positive, then the second weight function will have to be negative sometimes and positive sometimes so that the dot product is equal to zero. integral of w1(t)*w2(t) =0.

[xondiego]

Thanks for the videos!!, where/how do you get/compute w1(t) is like the eigenfunction, similar idea to the eigen vector of discrete case? what would be the eigen value related to the eigenfunction? is it the normal lambda? how this one is computed? in the discrete case the inner product is like xT dot w, but what does it mean ? this to interpret the pcscore or fpcscore s.