VARIABILITY WITHIN SYSTEM MODELS - A CASE-STUDY

Forms of variation which arise within system models are considered, using as an example a model for energy requirements in growing cattle. Problems encountered include the need to summarize groups of lines and to satisfy logical constraints on values of variables. Effects of Gaussian variation on prediction of weight gain of cattle are found to be substantial, with biases of up to 20% of the deterministic model outcomes, and standard deviations ranging from 20% to 150%. In some cases the distribution of predicted values is far from Gaussian.