Factor Analysis

  A variant of statistical analysis of data close to principal component analysis, the difference being largely one of terminology. Factors are the eigenvectors of the dispersion matrix ( Principal Component Analysis), multiplied by the square root of the corresponding eigenvalue. One calls the elements of this transformed eigenvector the factor loadings; they show the amount by which each original variable contributes to the total variance , which is the eigenvalue (= the sum of squared factor loadings). Like in principal component analysis, it is customary in factor analysis to set some elements in the factor matrix (viz. the matrix of factor loadings) to zero, using some problem-specific thresholding; this rank analysis or quantization reduces the dimensionality of the problem.