Application of canonical transformation with missing values to multitrait evaluation of Jersey type

A multitrait animal model was used to calculate predicted transmitting ability and reliabilities for final score and 15 linear type traits of 225,632 US Jersey cows. Records were adjusted for age and stage of lactation before analysis. The model contained effects for interactions of herd and date scored; year scored, parity, and age; and herd and sire; effects of permanent environment and additive genetics were also included. Of the 381,511 records included, some observations were missing for final score (8%), body depth (43%), and teat length (33%). The evaluation system used a canonical transformation, included several random effects, and estimated missing values with each iteration. Inbreeding was considered in the computations. Convergence was achieved in approximately 50 rounds of iteration. Correlations between animal and sire model predicted transmitting ability ranged from 0.56 to 0.95 and generally were higher for bulls than for cows and for more recent birth years. Genetic trend was strongly positive for dairy form, final score, and rear udder traits (height and width) and negative for udder depth. For other traits, genetic trend was small. This methodology should improve the accuracy of genetic evaluations for type traits of US Jerseys.