Study of statistical properties of genetic correlation using bootstrap technique

Abstract

The formula of standard error of genetic correlation based on half-sib intra-class correlation is highly approximate and depends upon the estimates and standard errors of other genetic parameters, which themselves are estimated with low precision. This results in unreliable estimation of the standard error of genetic correlation. The present investigation is an attempt to study some of the statistical properties of this important genetic parameter using bootstrap technique. The results showed that in 95% of cases the underlying distribution of the genetic correlation is non-normal. The bootstrap estimates of genetic correlation are biased to the extent of 5-10%. The higher estimates of standard error obtained by this technique have revealed that the standard error is underestimated by traditional estimation procedure using approximate formulae, available in literature. As expected the confidence intervals obtained by the percentile method are shorter in majority of cases as compared to confidence intervals obtained under normality assumption.