Random regressions to model phenotypic variation in monthly weights of Australian beef cows

Abstract

<p>Weights of beef cows recorded on a monthly basis were analysed using a random regression model. Data originated from a selection experiment in Western Australia, involving two herds of about 300 cows, Polled Herefords and a four breed synthetic, the so-called Wokalups. Weights were subject to large seasonal effects. Short mating periods and thus tight calving seasons led to substantial confounding between age and season at weighing. Records between 19 and 84 months were considered, up to 62 per cow, yielding 27728 and 29033 records for 922 and 1020 cows, respectively. Only phenotypic random regressions for animal effects, ignoring relationships, were considered. Covariances between regression coefficients and error variances were estimated by restricted maximum likelihood. A variety of models, involving random regressions on orthogonal polynomials of age, on segmented polynomials and on sine and cosine functions and different assumptions about the structure of error variances, were considered. Analyses identified a distinct cyclic, seasonal pattern of variation, both between animals and for temporary environmental effects. This could only partially be attributed to scale effects. Orthogonal polynomials proved well capable of modelling such sinuousity but required a high order of fit and thus a large number of parameters. Alternative curves utilising the known periodicity (12 months) provided more parsimonious parameterisations. Due to the high degree of confounding between age and season of recording their contributions to the total variance could not be separated.</p>