Power and Sample Size Calculations

What should you produce?

1. A written sample size justification. You need to describe how your sample is sufficient to address your research question. This includes a justification of your effect size. You also should include as an appendix your code (e.g., Stata) or GPower 3 Protocol.
2. A table and a figure that provides some indication of the sensitivity of your calculations. For example, how does your power change (or detectable difference or sample size) if you recruit 10% fewer subjects or if attrition is 20%? Show the results of your calculations for various scenarios so the reader can see how sensitive (and plausible) your analyses are.
3. 2-3 slides for your in-class presentation.

Steps to Take

1. Identify the effect size of interest for your study based on your analysis plan (nearly all will be a mean difference or correlation)
2. Pick an “anchor” effect size based on the literature. This can be influence by clinical or practical considerations. If you’re inclined to say that you don’t have enough information to pick an anchor or that power analysis asks for the impossible, then pump the brakes. “If you don’t know what you need, are you really ready to study the phenomenon?”
3. Use Stata or similar software to perform the analysis (including the sensitivity analysis).
4. Construct a table to show the results of the analyses.
5. Construct of a figure to show the results of the analyses.
6. Write up what you find, including your justifications for your assumptions.
7. Present to colleagues and get feedback.

Parts of the Workshop

Complete the relevant steps prior to the due date. Se Learning Suite for the due dates.

• Workshop 5a: Steps 1-5
• Worskhop 5b: Steps 6-7

Feedback to Colleagues

These are suggestions – they aren’t the only the things you should consider when evaluating your colleagues’ work. Be thoughtful and remember to provide actionable feedback.

1. Are the assumptions of the sample size calculations reasonable? Why or why not? What needs to be clarified?
2. Have they anticipated recruitment problems or attrition? Have they justified their choices of effect size or is more like “hand-waving”? Does the effect size they chose match up with their analysis plan and research question?
3. Does the table and figure communicate the key parts of their analysis? Does the table stand on its own? What else needs explaining in the table? How about the figure? Is it easy to understand? Is it ugly? Can it be improved to make it better looking and improve communication?

What will you turn into to me?

Turn in your completed power analysis (5a + 5b) that takes into account your colleagues’ feedback. You should also include the comments your received from colleagues as well your brief summary of how you dealt with their feedback.