MULTIPLE-TRAIT GENETIC EVALUATION FOR ONE POLYCHOTOMOUS TRAIT AND SEVERAL CONTINUOUS TRAITS WITH MISSING DATA AND UNEQUAL MODELS

A method for multiple-trait genetic evaluation for categorical and con tinuous traits was generalized to a polychotomous rather than a binary trait and to several continuous traits rather than one. Any missing d ata pattern was allowed. Breeding values were estimated based on an an imal model with fixed and random effects differing among traits. Equat ions in location parameters were solved iteratively within each Fisher scoring step. In each round of scoring, new solutions of the residual covariances among the categorical and the continuous traits were comp uted based on maximum likelihood estimation and used to reevaluate all partial regression coefficients of liability on the continuous traits for each missing data pattern. Simulation was used to assess the esti mation of the residual covariances. The other dispersion parameters we re treated as known because their estimation has been treated elsewher e and is analogous to restricted maximum likelihood.