A simple, rapid and reliable method for selecting or assessing the number of replicates for animal experiments

Abstract

A simple approach was developed for determining the number of replicates needed per treatment group to provide experiments of known power and sensitivity, where power equals the probability that a treatment effect would not go undetected if an effect existed and sensitivity equals the minimal treatment response that will be detectable. This approach, in turn, was used to construct reference tables, applicable across scientific disciplines, from which researchers may read replication requirements directly with ease, speed and reliability. To use the tables, one need only furnish a reliable estimate of the coefficient of variability expected among replicates, which may be obtained from prior observations on similar populations. The tabular data also enable a rapid, reliable assessment of the actual power and sensitivity of completed experiments, such as those contained within the published literature.