Statistical analysis of repeated measurements in physiology experiments

Experiments with repeated measurements represent the most common design in present use in physiology and endocrinology. They are also very common in animal nutrition and genetics. The common split plot for animals is described, in which animals remain on the same treatment during the course of the experiment and are measured repeatedly over time. Characteristics of these experiments are listed. Problems of correlated errors, heterogeneity in the error variance-covariance matrix and strategies for statistical analysis of classic split plot models are discussed, along with alternative approaches. Statistical analysis of areas under the curve is shown to provide little new information over standard analysis. Considerations of time as a continuous or discrete variable, orthogonal comparisons of treatment profiles or trends, and multivariate analysis of variance of data or regressions are covered. Bootstrap and permutation procedures and regression analysis of cumulative counts or measurements can provide additional insight into the biology of response to treatment in experiments with repeated measurements. Also discussed are special considerations necessary for analysis of binary data. (author). 25 refs, 3 figs, 3 tabs