Power analyses and sample size calculations are important parts of many research projects. Often, before data are even collected, it is necessary to calculate and justify the required sample size for a study. Choosing an appropriate sample size is important since it allows us to detect anticipated treatment effects and associations with appropriately high probability, or synonymously, high statistical power. If we have too few samples, the chance of overlooking existing effects is high. Drawing too many samples leads to wasted time and resources.
Sample size and power calculations rely on a few important quantities: effect size, desired power, sample size, and type I error (i.e., false positive) rate. For a given analysis plan, specification of any three of these quantities enables calculation of the fourth. For example, if a researcher wants a 90% chance to detect a certain correlation with a 5% type I error rate, the sample size that achieves these conditions can be obtained from a variety of widely available software packages.
In this short course we will define each component of these analyses and guide attendees on how to find the information necessary to complete a sample size calculation. Then, we will explore how to apply this information within the scope of several common experimental setups (one- and two- sample proportions and means, correlations, one-way ANOVA, etc.). Finally, we will use JMP and G*Power (a free software download) to practice obtaining required sample sizes based on different effect sizes and desired levels of power.
Becoming comfortable with sample size calculations can be a major asset for anyone who participates in research. After this short course, attendees will be able to explore power, effect size, and sample size within the context of their own projects in order to become more effective researcher.
Figure 1: JMP Sample Size Calculator
Due to technical difficulties there is no video recording available for this session.