|
KMID : 1025520010430010009
|
|
Journal of Animal Science and Technology
2001 Volume.43 No. 1 p.9 ~ p.22
|
|
|
Bayesian Analysis of Carcass Traits using Multivariate Threshold Animal Models and Gibbs Sampling with Missing
|
|
Lee, D. H.
Misztal, I./Bertrand, J. K.
|
|
|
Abstract
|
|
|
Bayesian inferences using Gibbs sampling from multiple traits animal threshold models with missing records were used to estimate genetic parameters for carcass traits with one discrete data (MS) and four continuous data (CW, DP, LMA, and BF) in Korean cattle (Hanwoo). Those results were compared to the estimates by REML that were gotten at previous study. Another purpose in this study was to compare estimates between age constant and weight constant as end point to make decision to slaughter. To implement Bayesian inference using Gibbs sampling in a threshold model, one threshold was constrained and reparameterization was needed. It is suggested that resealed threshold estimates could be used by posterior distribution of residual variance for categorical trait. Resealed threshold estimates would converge faster. The FCPD for missing residuals could be generated by ordering and, furthermore, Gibbs sampler for residual of underlying variable associated with missing categorical trait should be firstly obtained from FCPD of prior distribution from known properties. Gibbs samples for residual covariance components between categorical trait and linear traits should be obtained from FCPD having the property of conditional inverted Wishart distribution.
The point estimates for h²from marginal posterior distribution in this study were larger than those by REML. Otherwise; point estimates of genetic correlations were similar to those by REML. There were no indications of large differences is h²estimates between age- and weight-adjusted models in this study even though the same trait between different models were shown different estimates as pairwise comparison. h²estimates under the underlying scale on marbling score by Bayesian inference were 0.425 on age-constant basis and 0.409 on weight-constant basis in Korean cattle. Another method may be developed to apply multiple traits threshold model with missing records.
|
|
|
KEYWORD
|
|
|
|
|
|
FullTexts / Linksout information
|
|
|
|
|
|
Listed journal information
|
|
|
|