Short communication: Development and evaluation of predictive models of body weight for crossbred Holstein-Zebu dairy heifers

Abstract

Equations to predict body weight (BW) of crossbred Holstein-Zebu dairy heifers were developed and compared with current models (Heinrichs et al. for Holsteins, United States; Reis et al. for crossbred Holstein-Zebu, Brazil). The data set was constructed from 150 measurements of BW (320±107kg) and biometric measurements such as heart girth (HG, 161±19.5cm), withers height (WH, 126±11.0cm), and hip height (HH, 132±11.3cm) of heifers from 5 commercial dairy producers in the southern Amazon region in Brazil. The data were evaluated using mixed nonlinear models with herd as a random effect. Three nonlinear equations were fitted: BW (kg)=0.00058HG (cm)2.6135; BW (kg)=0.000618HG (cm)2.7362; and BW (kg)=0.000196HH 2.8793. An independent database was constructed to evaluate the models from 38 treatment means of 4 feeding trials: BW 258±54.3kg, HG 142.5±11.8cm, WH 113.2±6.0cm, and HH 118.7±9.1cm (mean ± SD). The evaluations were based on the relationship between observed and predicted values of BW by linear regression, root mean square prediction error (RMSPE), and concordance correlation coefficient analysis. Only the proposed model using HG accurately predicted observed BW, with bias (observed - predicted) of 4.83kg and RMSPE of 5.41% of observed BW (87.7% of random error). The models using WH and HH failed to accurately predict observed BW, with a bias of -3.06 and 72.02kg, and RMSPE of 9.40% of observed BW (75.2% of random error and 23.1% of systematic error) and 30.81% of observed BW (81.2% of mean bias). Additionally, the models of Heinrichs and Reis used for comparison did not predict BW accurately, with a bias of 19.32 and 29.37kg and RMSPE of 9.08% of observed BW (68.4% of mean bias and 31.4% of random error) and 12.58% of observed BW (81.9% of mean bias). The largest concordance correlation coefficient of the proposed HG-nonlinear model (0.930), compared with the models of Heinrichs and Reis of 0.845 and 0.708, confirmed the greater accuracy and precision of the new equation to predict BW in crossbred Holstein-Zebu dairy heifers. © 2013 American Dairy Science Association.