Development of a Stochastic Pig Compositional Growth Model

Abstract

A stochastic pig compositional growth model was developed using mixed model nonlinear functions. Serial BW measurements were fitted to mixed model nonlinear equations with three parameters and two random effects. Empty body protein mass (EBPRO) data were predicted from serial real-time ultrasound and BW measurements. Predicted EBPRO data were fitted to a nonlinear function of predicted BW (PBW) with one random effect: EBPRO = C (¦(PBW)) + cpi (¦(PBW))D, where ¦(PBW) = (1 - exp (b0 + b1 PBW + b2 (PBW)2), C and D are fixed parameters, and cpi is a random effect. The model also accounts for the relationship among the random effects for BW growth and cpi. Daily lipid accretion was predicted from genetic population-sex specific relationships between BW, EBPRO, and empty body lipid mass: empty BW = 0.93 BW and empty BW = a1 EBPROb1 + a2 (empty body lipid mass)b2. The model predicts a BW growth curve and daily compositional growth rate for carcass fat-free lean, carcass fat tissue, EBPRO, and empty body lipid. To reproduce the total variance, the residual variance of each variable was produced by multiplying the residual standard deviation of each prediction equation by a value sampled from a standard normal distribution. The stochastic model can be used to develop strategies to target a specified mean and distribution of carcass composition end points.