Maximum Likelihood Estimation of Genetic Components of Variation

Publisher Website
Google Scholar
Abstract
Maximum likelihood estimation of components of variation in genetical experiments is compared with least squares analysis. The extra labor of iteration in the former method is offset by its ability to handle a wide range of variance magnitudes. The latter method over-estimates the errors of components derived from smaller variances and under-esti-mates those derived from larger variances. An experimental design more suited to the easier least squares analysis is to have equal degrees of freedom for all variances and covariances in segregating generations and to determine the relative degrees of freedom in non-segregating generations by the heritability. Families contain only one or two plants and randomization is by plants and not by family plots.

This publication has 1 reference indexed in Scilit: