Summary

This case study reports on a goat feeding experiment that went wrong and attempts to rectify it. As the experiment was run in batches it was possible to use data that had already been collected to decide on an appropriate course of action. Various options were possible and the sample size estimation formula is used to evaluate the likely outcomes of each alternative.

The case study provides a good example of the art of making compromises between what is statistically desirable and experimentally possible. Another interesting feature of the design used for this study is the inclusion of a pre-treatment period when milk yield was recorded for use as a covariate in the statistical analysis.