Statistics & Data
Shapiro–Wilk gives p < 0.05 and many researchers jump to Mann–Whitney. But normality tests usually answer the wrong question, and Mann–Whitney does not compare
Statistics & Data
A meta-analysis turns 20 small studies into one estimate, or pools apples with oranges and reports a very confident number. Weights, fixed versus random effects
Statistics & Data
A significant result says nothing about how big the effect is. Cohen’s d and Hedges’ g, r and R², eta squared, odds versus risk ratios, reading small, medium an
Research Methods
A teacher tries a new method with one class and scores rise by 15 percentage points, so the method works. There are at least five other explanations. Here is ho
Research Methods
The number 384 appears in thousands of theses, yet few authors can say where it comes from. Here is how sample size really depends on your question, with hand-w
Research Integrity & Open Science
Most p-hacking comes not from fraud but from small choices that seem reasonable on their own: a few more participants, one variable dropped, a different outlier
