A multi-attribute utility model to optimise sow replacement decisions

An existing stochastic dynamic programming model is expanded so that its ou tput can be used as input to a multi-attribute utility model to determine t he optimal replacement policy in swine breeding herds. The models include s ix attributes: future profitability of the sow (from the dynamic programmin g model), piglet mortality I ate, pigs crushed by the dam, savaged and stil lborn pigs, undersized pigs at birth, and undersized pigs at weaning. The a im is to rank sows for replacement purposes. The model provides stable rank ings that differ from those based only on the dynamic programming model. Th ere was good correspondence between the model's ranking of 30 sows and rank ings provided by two experts.