|VAN TASSELL CURT|
|VAN VLECK L DALE|
Submitted to: Journal of Animal Science Supplement
Publication Type: Abstract Only
Publication Acceptance Date: 3/10/1995
Publication Date: N/A
Technical Abstract: A set of flexible Fortran programs is being developed for estimation of variance components with general animal models using Gibbs sampling. The program interface is similar to that used in the MTDFREML programs. The programs support multiple-trait models with an arbitrary number of co- variates, fixed effects, second animal genetic effects, and random effects for each trait. The programs manage data with any combination of missing observations. Genetic effects and independent random effects related to animal effects (e.g., direct or maternal permanent environmental effects) are generated as a block to increase the rate of convergence. The programs generate information for estimating variance components, variance ratios, heritabilities, fixed and random effects, and contrasts. In order to guarantee valid parameter estimates, proper Bayesian prior distributions are required for all variance components. The programs generate the Gibbs samaples but will not perform burn-in or convergence analysis or generate posterior distributions. Therefore, the user is expected to have an understanding of these aspects of Gibbs sampling. Some of these features may be added over time, as better algorithms are developed.