Selection of Fixed Effects in Sire Evaluation Models Using Akaike's Bayesian Information Criterion.

                                                   YASUHIKO WADA
                                                    AKIRA TAKEBE
                           National Institute of Animal Industry
                                             Tsukuba Norindanchi
                                              Ibaraki 305, Japan

                                              NOBUHISA KASHIWAGI
                        The Institute of Statistical Mathematics
                                         Minami Azabu, Minato-ku
                                                Tokyo 106, Japan

                                                       TAKEO ABE
                      Livestock Improvement Association of Japan
                                                       Koutou-ku
                                                Tokyo 134, Japan

ABSTRACT

  The authors proposed the method for selecting fixed effects 
in sire evaluation models using Akaike's Bayesian information
criterion (ABIC). This method was applied to the field progeny
testing records of 7288 Japanese Black steers selected from
the records collected of the field progeny testing program for
performance of beef production between 1978 to 1982. The analyzed
traits were Growth rate (kg/day), Carcass growth rate (kg/day),
Dressing percentage, Marbling score, Carcass grade, Carcass
price per weight (yen/kg). The fixed effects considered were 
Sire born prefecture, Calf born year, Claf born prefecture, 
Fatted prefecture, Slaughter house and Age class at slaughter.
By calculating ABIC's for 15 alternative models, the model
which included the combined fixed effect of {Calf born year} - 
{Slaughter house} - {Age class at slaughter} was selected 
as the best fit model for the five traits except Carcass
price per weight, and the model which included the combined 
fixed effect of {Calf born year} - {Fatted prefecture} - 
{Age class at slaughter} was selected for Carcass price per
weight. Slaughter house and Fatted prefecture were not included 
in one model at the same time, because these effects were 
strongly confounded. Sire born prefecture and Calf born
prefecture were not employed in the best fit models. These
results suggested that combined fixed effects were useful
effects of the model in the Japanese Black sire evaluation.
In the future, Fatting method which is not confounded with
Slaughter house should be considered in stead of Fatted 
prefecture.

Key words

                 Anim. Sci. Technol.(Jpn.) 64(7): 678-685. 1993.