K.R. Koots, J.P. Gibson and J.W. Wilton
Anim. Breed. Abst. (1994) 62:825-853
Abstract
Phenotypic and genetic correlation estimates among beef cattle traits were taken from an extensive search of the animal breeding literature from 1940 to 1991. Unweighted means among 38 traits are presented; estimates were missing or poorly estimated for many trait combinations, including important groups, such as those between reproductive and carcass quality traits. Weighted mean correlations are presented for a subset of 21 traits, again with many trait combinations missing. A weighted least-squares analysis of estimates of each correlation showed several factors significantly (P < 0.10) affecting the estimates, including breed, country, sex and decade in which data were collected. Other factors [data origin (field data or experimental data), feeding regime (range or feedlot) and estimation method] generally did not significantly affect genetic and phenotypic correlations. Fixed effect models accounted for little of the observed variance of estimates between data sets, which were about four times the predicted error variance derived from standard formulae. Genetic correlation estimates were correlated with corresponding heritability estimates and phenotypic correlation estimates, and with other genetic correlation estimates derived from the same data set, more often than expected by random chance. The implications of these associations among parameters are discussed in the context of obtaining positive definite variance-covariance matrices required in breeding programmes. The weighted mean genetic and phenotypic correlations presented here for growth traits are recommended for genetic evaluation programmes in most cases. If parameter estimates are available within a population, these local estimates could be combined with the literature averages reported here, thereby using all the available published information for each trait combination. Further estimates of correlation among many trait combinations would add little to existing knowledge, but there are substantial gaps in the estimates required for planning comprehensive improvement programmes.