A Note on Matrix Inversion by the Square Root Method
David G. Durand · Journal of the American Statistical Association · 1956
rT HE square root method, also known as Choleski's method in the nonstatistical literature, has been discussed recently by Duncan and Kenney [1, pp. 16-29], Dwyer [4; 5, Sec. 6.5, 13.4, 13.7, 13.8], Fox, et al. [7, p. 159], and others, with a bibliography in [5, p. 118]. The essence of the method is the reduction of a symmetric matrix A to the product S'S-in which S is a triangular matrix with zeros below the main diagonal, and S' is its transpose. Then, if A and hence S is non-singular, A-' can be obtained either by inverting S and utilizing the relation