Interpreting experiments on egg production--statistical considerations

A given data set can be analyzed many ways, but only one is the correct analysis based on the design actually used when running the experiment. This work gives a tutorial-like illustration of the effects of the presence of a regression variable (or covariate) on the recorded responses in an experiment set up as a standard factorial design and shows how the analysis results are to be adjusted for the presence of covariates. An underlying assumption of a factorial model is that each of the treatments (e.g., diets) is randomly allocated to different subjects (hens). When many measurements (e.g., over time) are made on the same subject (hen), this independence assumption is violated; in these cases, the design is an example from the class of repeated measures designs. The difference in analysis between factorial designs and repeated measures designs is also discussed. Then, the 2 concepts are merged wherein the results for a repeated measures analysis have to be adjusted for the presence of covariates. The paper concludes with analyses on the results of egg production responses from an experiment in which repeated measurements were made on the same hens and in which an unanticipated temperature covariate was present.