A note on the use of the analysis of covariance to reduce standard errors in experiments on milk production

Abstract

The use of changeover designs and of covariance analysis in experiments on lactating cows is reviewed. Continuous treatment designs are criticized because of their large coefficients of variation, e.g. of the order of 20 % for milk yield. Although changeover designs have smaller errors they are criticized because the results have strictly limited practical application. Covariance analysis applied to the data from continuous treatment designs offers a means of increasing their precision. Measurements made in a preliminary period may be used to adjust experimental measurements made subsequently for animal to animal variation. The covariance method was tested on data from thirtytwo heifers in an experiment involving factorial sequences of treatments. Weeks 5-12 of lactation formed the control period and weeks 13-22 the experimental period.