Nonlinear Mixed Effects Model for Swine Growth

A mixed effects model version of a common swine growth function is introduced. This version, in which the mature BW of each pig is considered random, accounts for increasing variation with age and serial correlations across age that are common in serial BW growth data sets, thereby providing a much better fit and significantly smaller parameter SE. The inclusion of random effects also reduces the impact of selective sampling, which occurs when fast-growing pigs are removed for marketing, and provides a statistic (i.e., variance estimate) that describes the between-pig variation within a group. Comparisons between the fixed and mixed effects models are made using swine growth data consisting of 93 pigs with biweekly BW data from 54 to 138 d of age. In addition to a much better fit, the mixed effects model version is easily adaptable to stochastic modeling because only an additional random effect (i.e., mature BW) for each pig needs to be generated.