Even when the research questions and study design seem straightforward, the ensuing power analysis can seem technically daunting. It is often helpful to break the process down into five components:

• Study Design: What is the structure of the planned design? This must be clearly and completely specified. What groups and treatments ("cells" and "factors" of the design) are going to be assessed, and what will be the relative sizes of those cells? How is each case going to be studied—that is, what is the primary outcome measure ("dependent variable")? Will covariates be measured and included in the statistical model?

• Scenario Model: What are your beliefs about patterns in the data? Imagine that you had unlimited time and resources to execute the study design, so that you could gather an "infinite data set." Characterize that infinite data set as best you can using a mathematical model, realizing that it will be a simplification of reality. Alternatively, as is common with complex linear models, you may decide to construct an "exemplary" data set that mimics the infinite data set. However you do this, your scenario model should capture the key features of the study design and the main relationships among the primary outcome variables and study factors.

• Effects and Variability: What exactly are the "signals and noises" in the patterns you suspect? Set specific values for the parameters of your scenario model, keeping at most one unspecified. It is often enlightening to consider a variety of realistic possibilities for the key values by performing a sensitivity analysis, to explore the consequences of competing views on what the infinite data set might look like.

• Statistical Method: How will you cast your model in statistical terms and conduct the eventual data analysis? Define the statistical models and procedures that will be used to embody the study design and estimate and/or test the effects central to the research question. What significance levels will be used? Will one- or two-sided tests be used?

• Aim of Assessment: Finally, what needs to be determined in the power analysis? Most often you want to examine the statistical powers obtained across the various scenarios for the effects, variability, alternative varieties of the statistical procedures to be used, and the feasible total sample sizes. Sometimes the goal is to find sample size values that provide given levels of power, say 85%, 90%, or 95%.