Buy me a coffeeFive different “Cohen’s d” statistics for within-subject designs
What is this?
Excerpt from blog: What I’ll show here is that there are at least 5 different and non-equivalent ways that people might compute a d-like effect size (which they would invariably simply call “Cohen’s d”) for Jeff’s dataset, and the resulting effect sizes range from about 0.25 to 1.91. I’ll compare and contrast these procedures and ultimately choose one that I think is the least crazy, if you must compute a standardized effect size (more on that later).