Empirical Sampling Estimates of Genetic Correlations

Abstract

A method of generating samples from a normally distributed population is described which has the advantage of being parameter free. Sample coefficients can be developed with zero means and unit variances and then different sets of parameter values used with the sample coefficients to develop sample sums of squares and crossproducts from populations with different parameters. This method is used to construct sample estimates of genetic correlations for 3 sample sizes and 24 sets of parameter values. The sampling variances of these estimates are compared with those expected by application of Reeve''s (1955) formulae.