Choose Your Own Data-Analytic Adventure

bob.8
So how do I determine my statistical power and sample size?

First, you don't get to determine both. Statistical power is a function of sample size (and the population's correlation and your alpha level). Sample size is a function of statistical power (and the population's correlation and your alpha level). In general, there are four components to statistical power analysis, and if you know three components (any three), the fourth is mathematically locked in:

  • Effect Size (i.e., the w statistic)
  • Sample Size
  • Alpha Level (usually .05)
  • Statistical Power (or 1 - Beta Level)
You choose one of the components to calculate based on your best guesses for the other three components. So which do you want to calculate?

I don't know. Which do you suggest?


You reply, after noting that you should stick with the conventional alpha level of .05 to avoid suspicions of tinkering, and after noting that effect size (known as w) is nearly impossible to interpret unless you calculate it by hand, so we won't calculate alpha or w:
Let's calculate the sample size you should collect for your study.
or
Let's calculate the statistical power of your study.