Estimation of Genetic (Co)variance Components for International Evaluation of Dairy Bulls

Publisher Website
Google Scholar
Abstract
Within- and across-country genetic parameters need to be estimated prior to an international genetic evaluation of dairy bulls. A procedure based on the Expectation Maximization algorithm to produce restricted maximum likelihood estimates of such parameters, using national evaluation results from different countries, was tested by simulation. Individual performance records were generated for two populations of dairy cattle with separate breeding programmes but considerable genetic exchange to create ties between the populations. A genotype-by-country interaction effect was generated by simulating the data according to a genetic correlation of 90 between performance in the two countries. Within-country national evaluations were computed with animal models. The estimation procedure was tested using bull national evaluation results. The impact of factors such as data connectedness, time period, and bias in national evaluations on estimated parameters was investigated. When all data were included in the analysis, correct estimates of genetic parameters were obtained. When older data were excluded, within-country variance estimates were biased downwards. When national evaluations of imported bulls, often considered biased owing to preferential treatment of daughters, were excluded, the genetic correlation was underestimated by 12%. When 5-15% bias was introduced into the evaluations of imported bulls in one country and the latter were included in the estimation procedure, genetic correlations were only slightly underestimated. When genetic links between populations were weakened, the genetic correlation was seriously underestimated. A better estimate was then obtained based only on a well-connected subset of the data instead of the entire dataset. The procedure was tested on real data from five countries: Germany, France, Italy, the Netherlands, and the United States of America. Genetic correlation estimates between production traits in these countries ranged from .90 to .97.