G*Power

Suggested Exercise
(~10 minutes)

Assume you’re a biologist conducting a study to determine if a new diet supplement has any significant impact on weight loss in mice. Your hypothesis is that mice receiving the diet supplement will lose more weight than those receiving a placebo.

Introduction
Open G*Power and select the “Test family” dropdown menu. Choose “t tests” and then “Means: Difference between two independent means (two groups)” from the “Statistical test” dropdown. (Why this statistical test?)

Define Input Parameters
In the “Type of power analysis” dropdown menu, select “A priori: Compute required sample size – given α, power, and effect size”.
Set α as 0.05. (Why 0.05?)
Set Power (1-beta error probability) as 0.80.
Finally, the effect size needs to be defined. For the purposes of this exercise, you can use a medium effect size (d = 0.5) as a standard measure. However, in actual studies, this should be derived from literature reviews or preliminary studies.

Perform Analysis
Click “Calculate” to obtain the required sample size.
Consider the results. The output gives the total sample size required for the study, which you’d generally divide equally among your two groups (the supplement and placebo groups).

Recompute Analysis
Now, let’s assume you only have 30 mice available for the study. Let’s see what our power will be given these constraints.
Select “Post hoc: Compute achieved power – given α, sample size, and effect size” from the “Type of power analysis” dropdown.
Enter the total sample size as 30 and calculate. Consider the achieved power.

What do you suppose the potential consequences would be if the power is below the desired 0.80?

Helpful Literature
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2996198