Slides

Bland (2009) has commented that funding agencies and journals put investigators in an inconsistent position: Funding agencies ask for statistical power calculations to test one hypothesis for the primary outcome, yet journal editors ask for confidence intervals. Many funding agencies, however, rely on the conventional power calculations advocated by most trialists. Therefore, this presentation primarily focuses on power calculations and adapts conventional advice from trialists to nonrandomized or observational studies because they introduce complexities that randomized trials do not need to consider. For example, investigators may not be able to estimate the power or precision of their proposed comparisons until they have generated the propensity score and constructed matched cohorts, which may exclude patients and interventions that appeared eligible when the cohort was assembled.