Planning an experiment in the company of measurement error

Abstract

"Textbook" calculations of statistical power and/or sample size follow from formulas that assume that the variables under consideration are measured without error. However, in the "real world" of behavioral research, errors of measurement cannot be neglected. A recent sample-size determination approach is easily adapted to incorporate unreliability information for both completely randomized and randomized block analysis-of-variance designs. A worked example presents an instance wherein a blocking strategy is clearly advantageous assuming infallible measuring instruments, but not when the same instruments are granted fallibility.