Quality, innovations and statistics · 2 steps
Kappa and Bland-Altman
Percent agreement and correlation can overstate agreement: kappa corrects categorical readings for chance, and Bland-Altman shows systematic and proportional differences in continuous measurements.
Left: two readers' calls on 100 studies build up in a 2×2 table with P₀, Pₑ and kappa; then a scatter plot of two measurement methods with the correlation coefficient. Right: kappa on the Landis-Koch scale with nuclear medicine examples, and the Bland-Altman plot with mean difference and limits of agreement.
Percent agreement and correlation can overstate agreement: kappa corrects categorical readings for chance, and Bland-Altman shows systematic and proportional differences in continuous measurements.
Source: kappa formula, Landis-Koch classes, Deauville and V/Q examples and the Bland-Altman definition from the book (29.5). 2×2 tables and SUV measurements are representative (measurement error SD 0.3, fixed difference +0.6, proportional difference 12%).
How to use: a single click on the drawing pauses and resumes; ↑ ↓ or the buttons below change the scenario; use the time bar below to go to any moment; ← → change steps.
Presentation mode ↗The animation alone, sized for full screen; for projecting in lectures and talks. ← → step, space plays/pauses.
Embed
Paste this code into your own page (course site, slides, learning platform). The animation arrives working, with its source written underneath.
In the book
In the book this animation runs in place of these figures: