Power Analysis

Abstract

Abstract Power analysis is a tool for investigating trade‐offs between error rates, sample size, and precision, and has several different connotations depending on the underlying problem or objective. In the hypothesis testing context, power analysis can be used to investigate trade‐offs between Type I and Type II errors or to investigate the effect of varying sample size on statistical power. Power analysis has a broader definition in the context of environmental statistics, however, being widely used by statisticians to examine trade‐offs when designing sampling programs. In this context, power analysis largely refers to the allocation of survey effort in relation to some desired level of precision. The investigator might compare the effects of different design options (e.g., simple random sampling, stratified random sampling) or different types of estimators (model‐based estimators versus design‐based estimators). Power analysis is best viewed as a prospective tool for optimizing study design to meet objectives with a given set of constraints. In contrast, retrospective power analyses, whereby researchers fail to achieve a “significant” result and argue that it was likely due to insufficient sample size, are flawed and should be avoided.