Optimizations with uncertainty in pork production

The objective of this thesis was to identify areas for optimization in swine production. The first four investigations pertain to voluntary removal and replacement at the sow herd level. The fifth investigation pertains to production from weaning to slaughter.;The first objective was to estimate the validity of historical and current litter size as a predictor of future litter size in three herds. Significant predictors which were farm-specific included previous litter liveborn, parity, litter service number, farrow month and a random effect of sow. Model prediction improved as the numbers of parities in the model increased. Best Linear Unbiased Prediction estimates of eight-litter lifetime average liveborn were relatively effective in predicting the odds that the lifetime average would equal or exceed nine pigs and improved as the number of parities in the model increased.;The second objective was to use a retrospective case-control study to estimate the ability of three herds to replace animals removed for reasons attributed to poor fertility, poor fecundity, and old age, with more productive females. The success of removal/replacement events differed across herds and across removal reasons.;The third objective was to estimate the ability of three herds to replace animals that were poorly productive animals with more productive females according to contemporary herd performance and use a partial budget to estimate the implications of performance-based removal/replacement strategies. The success of removal/replacement events differed across herds, herd contexts and removal reasons. The financial model suggested that voluntary removal/replacement did not add herd financial or productivity gains.;The fourth objective was to develop a model to determine optimal replacement time for sows according to the net present value of retaining the sow in space compared to the net present value of immediate replacement with a gilt of average productivity and average longevity. The model uses litter liveborn to estimate the current and future productivity of a sow in the herd.;The fifth objective was to assess the effectiveness of Taguchi Methods for use as a tool to identify optimal factor settings to achieve system optimization. Optimal characteristics of nursery groups at placement were identified which minimized the cumulative deads, culls and lightweight pigs at the end of the finishing phase. Taguchi Method analysis and analysis of variance both identified the same nursery-level factors [and factor levels] associated with differences in the total proportions of deads, culls, and lightweights pigs at the end of finishing.