Some Properties Of Three Factor Contribution Matrices

Some properties of three matrices which have been proposed as quantitative bases for factor interpretation are discussed. The first of these contains as elements the covariances of the tests with weighted factor scores. Interpretations of the row sums and column sums of this matrix are given and its independence of the primary factor-reference vector distinction is demonstrated. Two matrices of part correlations are shown to be proportional by columns to the primary factor pattern (and thus to the reference vector structure). One of these matrices is shown to be equal to the reference vector structure.