Assuming we have an appropriate variance model, there is a problem associated variance component matrices which is independent of the algorithm used to obtain that sampling variation in the data often supports a non-positive definite components, ie. 1995) matrix and is generally quite efficient. ASREML uses a quadratic convergence method based on the Average Information (Gilmour et al. 1971) is now the preferred method and is implemented in many programs including ASREML (Gilmour et al. These include genetic and residual matrices between traits and between times which are used to calculate Restricted Maximum Likelihood (Patterson and Thompson heritabilities and genetic correlations.
Keywords: Genetic correlation estimation, average information, REML, variance components INTRODUCTION Animal breeders are often interested in estimating variance/co-variance matrices from data. Difficulty arises when the assumed model is not correct and because of sampling variation in the matrices. Gilmour NSW Agriculture, Orange Agricultural Institute, Forest Road, Orange, NSW 2800 SUMMARY The paper explains why it is often difficult to estimate co-variance matrices from data and describes the variance structures available in ASREML to investigate the problems. Vol13 VARIANCE STRUCTURES AVAILABLE IN ASREML A.
0 kommentar(er)
