USMLE purpose: Distinguish Cohen’s kappa, Kendall’s tau, and Pearson correlation coefficient when interpreting agreement and correlation questions.

How to use this page

  1. Decide whether the question is about agreement, ordinal association, or linear correlation.
  2. Match the statistic to the data type.
  3. Remember that correlation does not imply causation.
  4. Use the rapid recall toggles before the practice questions.

One-liner

Use Cohen’s kappa for agreement between raters, Kendall’s tau for ordinal/ranked association, and Pearson r for linear correlation between continuous variables.

Statistic selection table

Statistic Best use Board-style clue
Cohen’s kappa Inter-rater agreement beyond chance Two clinicians/radiologists agree on diagnosis
Kendall’s tau Ordinal/ranked association Rank order, nonparametric ordinal data
Pearson correlation coefficient Linear correlation between continuous variables Scatterplot, r value, straight-line relationship

Pearson r interpretation

r value Meaning
+1 Perfect positive linear correlation
0 No linear correlation
-1 Perfect negative linear correlation

Anki-style rapid recall

Practice Question

Two radiologists independently classify chest x-rays as pneumonia present or absent. Which statistic is most appropriate to measure agreement?