I aim to find which directions in my data have “vary greatly”. To do that, I understand a method called PCA, which uses the eigenvectors of the covariant matrix to find them.