Statistical power and research in occupational therapy

The rationale of statistical hypothesis testing is reviewed, and the concepts of experimental error and power are introduced and discussed. A post hoc power analysis of 205 statistical tests from 22 articles reporting occupational therapy research is presented. The effect sizes and power coefficients for each of the statistical tests reviewed were analyzed, and the evaluation revealed a significant difference in power values based on effect-size calculations: Studies reporting small and medium effect sizes were associated with low power and were subject to a high probability of Type II experimental error. The power of the tests reviewed and the estimated Type II error rates were shown to vary significantly for small, medium, and large effects. This was not true for Type I error rates, which remained constant for all effect sizes. The implications of the findings and the importance of pre-experimental power evaluations for fields such as occupational therapy, where research endeavors frequently reveal small effects, are discussed.