From one pig to a group of pigs: accounting for links among individual parameters in growth modeling at the population scale

Several pig growth models exist, but these models do not account for variation among pigs. Because the nutrient requirement of an average pig is not representative of the requirement of a group of pigs, a feed formulated upon the requirement of the average animal within a batch will not fit animals which requirements are above the mean. Knowledge of the variation and covariation of key model parameters is essential to simulate the response of a group. The objective of this study was to quantify these traits for model parameters describing feed intake and growth, based on observations obtained for individual pigs. We performed a dual multiple factorial analysis on a dataset of 1028 pigs originating from 32 different groups studied in the IFIP experimental station (Romillé, 35). Each individual pig was characterized by five model parameters; three of them described growth and two of them described ad libitum feed intake level. The distributions of parameters per population were summarized as a vector containing the means, with a covariance matrix describing the variation around the means. Neither the mean vector nor the covariance matrix could be generalized among the 32 groups, indicating that the population structure differed among the groups. In practical situations, it is virtually impossible to obtain all information required to describe the population structure. To simulate the on - field response of a population, we therefore propose that the mean vector of parameters has to be obtained for each situation and that a common covariance matrix structure may be used.