Principal Component Analysis

Assume a 2-D Gaussian for the intrinsic distribution

• People all distributed in a gaussian cloud of height and weight

• Principal axis of variance is then "Size"

• Orthogonal axis would then be the "Overtall/Undertall" vector