PDMATRIX - programs to make matrices positive definite

John M. Henshall and Karin Meyer



Combining estimates of covariance components for a number of traits from various sources while ensuring the resulting covariance matrix is positive (semi-) definite, is a routine problem in multivariate genetic applications. Frequently, estimates available differ greatly in accuracy.

PDMATRIX consists of two programs to generate positive definite covariance matrices from individual estimates, accounting for differences in accuracy. FLBEND modifies a non-positive definite matrix comprised of covariance estimates for pairs or blocks of traits. ITSUMCOV deals with the specific case where we want to combine results from multiple, multivariate analyses of various subsets of all traits considered.

PDMATRIX is written in FORTRAN 90. Source code and compiled versions for Linux and Compaq Alpha stations are available by downloading from :


K.Meyer, 2002