Third International Congress on Quantitative Genetics, Hangzhou, China,
August 19-24, 2007.


Karin Meyer

Animal Genetics and Breeding Unit, University of New England, Armidale, NSW 2351, Australia

Modern analyses to estimate genetic parameters and covariance components generally fit a linear mixed model, and restricted maximum likelihood (REML) is a preferred method of estimation. WOMBAT is a program to carry out such analyses for continuous traits.

WOMBAT allows for multiple fixed effects and covariables, accommodating nested effects and a variety of basis functions, including splines. Random effects fitted can be distributed proportionally to the numerator relationship matrix (NRM) between subjects, an identity matrix or a general, user-defined covariance matrix. The NRM is determined within the program from a list of pedigrees for either the so-called animal model or a sire model.

WOMBAT accommodates multivariate analyses for numerous traits and readily allows for standard full rank estimation, as well as reduced rank estimation fitting only the leading principal components of selected covariance matrices. In addition, analysis of `function-valued' traits is accommodated through random regression models allowing for a variety of basis functions.

WOMBAT incorporates state-of-the-art numerical procedures to ensure computational efficiency and reliable maximisation of the likelihood function. Methods available include the `average information' and the `parameter expanded expectation maximisation' algorithm. WOMBAT is suitable for the analysis for very large data sets, and has been used for REML analyses involving close to 1 million equations in the mixed model. At convergence, estimates of covariance components and genetic parameters are provided together with their approximate lower bound sampling errors, and generalised least-squares estimates of fixed effects are given.

In addition, WOMBAT gives best linear unbiased predictions (BLUP) for random effects, and calculates residuals. Furthermore, for selected models WOMBAT can be used as a simple simulation program, replacing observations with simulated values from multivariate normal distributions.

WOMBAT is available (free of charge) by downloading from
Material provided consists of the compiled program for Linux operating systems (including the `CygWin' emulation under Windows) for 32- and 64-bit machines, a user manual and a suite of worked examples.