On Estimation of Genotypic Correlations and Their Standard Errors by Multivariate REML using the MIXED Procedure of the SAS System

This paper explains a simple strategy to develop code for multivariate (multitrait) mixed-model analysis using the MIXED procedure of the SAS System, when the corresponding univariate model has a simple variance components form. Such models are needed for estimating genotypic and phenotypic correlation among traits in plant breeding experiments, and they provide best linear unbiased predictions (BLUP) of genetic effects which may be more accurate than univariate BLUP. We give some hints how to speed up computations, and it is shown how to obtain asymptotic standard errors of correlation estimates with little effort.