Two‐step versus joint analysis of Von Bertalanffy function

Animal growth has been modelled using several mathematical functions (Logistic, Von Bertalanffy, Gompertz, etc.). These functions are described by a reduced number of parameters. In some cases, these parameters are assigned a biological interpretation (adult weight, maturing rate, etc.) and so the breeding goal can be closely related to the change of performance in these parameters, with the objective of altering the shape of the growth curve. In this sense, some studies have been focused on the estimation of (co)variance components of these parameters (B rown et al. 1976; F itzbugh 1976; D e N ise and B rinks 1985; K oenen and G roen 1996). These studies are based on a two‐step procedure. In the first step, a growth curve is fitted separately to the data of each individual animal, afterwards, a mixed model analysis is applied to obtain (co)variance components. In this second step the estimates of production function parameters from the previous step are taken as records. Recently, V arona et al. (1997, 1998) have described a Bayesian procedure which allows the particular parameters of any production function to be estimated jointly and the (co)variance components between them. This procedure provides a considerable advantage over Maximun Likelihood approaches to joint analysis (Z ucker et al. 1995) and makes use of all the available information. The aim of this study is to compare both procedures in a simulation scheme under the assumption of the Von Bertalanffy growth function.