For a given square matrix, a nonzero vector that changes length, but not direction, when multiplied by the matrix. The computation of principal components for a set of Variables uses the eigenvectors of the variables' Covariance or Correlation matrix.