PREDICTION OF BODY-WEIGHT FROM BIOMETRIC TRAITS IN TWO FOUNDATION LINES OF BROILER
Keywords:
Prediction, biometric traits, broilersAbstract
The research was conducted at the Poultry Breeding Unit, National Animal Production Research Institute (NAPRI), Shika, Zaria. A total of 1680 birds comprising of 1440 hens and 240 cockerels were used in 3 generations for the experiment. Mating ratio was 1 cockerel to six hens in each pen for different lines. Eggs were sorted sire wise before setting. Chicks were weighed individually at hatch, using a digital weighing balance and recording their individual body weights and wing tagged for easy identification. Weighing was done subsequently on a biweekly basis. Biometric traits: Neck length (NL) Back length (BL), Keel length (KL), BRLBreast length (BRL) and Thigh (TL) length were measured with a tape rule at 8 weeks. The deep litter system of management was used to rear the birds from day old to 8 weeks of age. Stepwise multiple regression procedure was used to find the best linear combination of metric variables that best explains the live weight. The regression analysis was carried out using the SASREG procedure of the SAS (2008) Package. The correlation coefficients of body weight and the linear body measurements were also determined. From the correlation matrix, data for the principal component factor analysis were generated. All predictors had positive correlations within generations except at base generation where negative association (r = -0.010) existed between breast and thigh length. The highest prediction power (R2 = 0.96 and 0.98 %) was obtained in the third generation for sire and dam line, respectively. Generation 1 recorded percentage variance of 41.70%, 29.60% and 20.8% respectively for PC1, PC2 and PC3. Generation 2 also had three components extracted with share variance accounting for 58.2, 29.6 and 21.0% of the total variance. Three components were also extracted in generation 3 with share variance of 71.2%, 48.9% and 22.6 % respectively. In concluding the low multicollinearity observed between body weight and body measurements implies that the prediction power obtained in this study is highly reliable for estimating body weight in settings where measuring scales are absent