The Estimation of Repeatability and Heritability from Records Subject to Culling

Abstract

The maximum likelihood estimation of repeatability from first and second lactation records of a herd subject to culling is discussed. The method is applicable only when it can be assumed that, had there been no culling, the variances of the first and the second lactation records would have been equal, and when all the first records are available whether or not the cow has a second record. The efficiency of the more usual estimate of repeatability, based on the regression of second records on first records, is shown to be low when the repeatability is high. In many cases the calculation of the maximum likelihood estimate would seem to be well worthwhile. The use of the method in estimating heritability is mentioned. All illustrative example is given.