Effect of sampling errors on efficiency of selection indices 1. Use of information from relatives for single trait improvement

Abstract

Summary: An analysis is undertaken of the effect of errors in estimates of parameters, particularly the intra-class correlations, on the response from selection for one trait using an index of individual together with full- and/or half-sib family records. A distinction is drawn between the response (R) possible with use of the optimum index, that predicted (R) and that achieved (R^*) with an index which uses the sample estimates of the parameter values.It is found that the loss of efficiency (R^*–R) using sample estimates is very small even for estimates far from the correct value. R is more sensitive to errors, particularly of the heritability and phenotypic variance estimates. Since the latter also appear in the prediction of response from individual selection, errors in predicting the relative responses from index and individual selection are small. Expected values of the proportional loss in response, are approximately proportional to the variance of the estimate of intra-class correlation. It is shown that in practice initial experiments with 20 or so families may be sufficient to get average proportional losses down to less than 1%.