Influence of sample size on precision of genetic correlations in red clover

Abstract

Previous investigations of genetic correlation (r(G)) indicate that r(G) is sensitive to sampling error. A nonparametric bootstrap method provides an alternative for estimating standard error and confidence intervals. Using the bootstrap method in a half-sib red clover (Trifolium pratense L.) population, we evaluated the effect of sample sizes on the point estimate, its sampling error, and confidence interval of r(G). We conducted separate analyses for varying numbers of replications, families, and plants per family using a seedling data set. Precision of r(G) was affected by sample sizes and traits measured. To achieve a reasonable standard error and a small confidence interval of r(G), a minimum of 42 half-sib families each having nine plants in four replications would be required. Although this study was limited to seedling traits in a greenhouse, similar results may occur when measuring other traits.