Modelling growth of young bulls
by
Vivienne Doogan
Qld Department of Primary Industries
Coauthors: Scott Foster (Qld Department of Primary Industries), Tony Swain (Qld Department of Primary Industries), Dick Holroyd (Qld Department of Primary Industries)

In a longitudinal study of young bulls in Central Queensland, measurements were recorded up to 10 times over a 12 month period on 29 Composite and 44 Brahman bulls. Bull age was 306 ± 4 (mean ± SD) days at the time of first measurement. The aim of the study was to compare the development of the two breeds over time.

For growth measurements such as liveweight and scrotal circumference, individual bulls followed their own pathway over time. Linear mixed models using residual maximum likelihood (REML) were used to analyse these growth variables. Two approaches were compared. Visual inspection of the pathways over time indicated they could be approximated by a quadratic polynomial. Therefore the first approach was to use random quadratic coefficients regression analysis. To allow for greater flexibility in the form of the pathways the second approach used cubic smoothing splines (Verbyla et al. 1999), fitting an overall spline, breed splines and individual bull random coefficients and splines. Results from the two methods of analysis will be presented and compared.

Verbyla, A.P., Cullis, B.R., Kenward, M.G. and Welham, S.J. (1999). The analysis of designed experiments and longitudinal data by using smoothing splines (with discussion). Appl. Statist. 48: 269-311.

Date received: August 28, 2002