Chapter 10 Power and sample size calculations

Power and sample size calculations are a common request for a GRA. In order to estimate the power or calculate the necessary sample size, you will need some measure of effect size. Often, the PI may not have an effect size in mind – in this case, offer some example scenarios and gauge the PI’s reaction. Examples:

• “If you saw that a subject being treated in group B had a 2 mm increase in the outcome, would you be impressed? Would that be a ‘big’ or notable change?”
• “How much of an impact would you need to see in order to catch you attention in an abstract? A 50% increase? Double the outcome?”

Your calculations for power/sample size could be done several ways in R: the pwr and pwr2 packages can be a good place to start. SAS offers the PROC POWER procedure for doing power and sample size calculations. I have often used both R and SAS functions for the same calculation and compared the results. Typically, SAS has much more in-depth documentation – if you are wondering about the theoretical details for a calculation, SAS documentation is a helpful resource.