SELECTION ON LEAN GROWTH IN A NUCLEUS OF LANDRACE PIGS - AN ANALYSIS USING GIBBS SAMPLING

Data from 4150 Landrace pigs tested during the period 1989-94 for back fat thickness and age at 100 kg in an open selection nucleus were anal ysed with the standard restricted maximum likelihood/best linear unbia sed prediction method and with a Bayesian approach based on the margin al posterior distributions of parameters of interest achieved via Gibb s sampling. Breeding values and fixed effects were sampled from normal distributions and (co)variance components from inverted Wishart distr ibutions. The Bayesian analysis indicated that the selection was effec tive for both traits. Assuming flat priors for the (co)variance compon ents, the posterior means of the annual rates of response to selection for both traits were -0.473 days and -0.212 mm. The influence of info rmative priors constructed from (co)variances estimated in the French Landrace breed on inferences about genetic and common environmental pa rameters, genetic group effects and total and annual responses was als o examined.