Graph description: Kappa statistics
The graph shows seven curves for seven different probabalities than an
observer will be correct.
The horizontal axis shows probablility of a true yes from zero to one.
The vertical axis shows the predicted kappa, also on a scale from zero to one.
Each curve starts at probablility of a true yes = zero and predicted kappa = zero
and rises to a maximum kappa at probablility of a true yes = one half,
then falls symmatrically to predicted kappa = zero at probablility of
a true yes = one.
The curve for 99% chance correct rises steeply until probablility
of a true yes = 0.1, then rises slowly to maximum kappa = 0.95 approximately,
representing very good agreement between probablility of a true yes = 0.05
and 0.95.
The curve for 95% chance correct rises steeply until probablility
of a true yes = 0.2, then rises slowly to maximum kappa = 0.8 approximately,
representing good agreement between probablility of a true yes = 0.1
and 0.9.
The curve for 90% chance correct rises steeply until probablility
of a true yes = 0.3, then rises slowly to maximum kappa = 0.65 approximately,
representing moderate agreement between probablility of a true yes = 0.1
and 0.9.
The curve for 80% chance correct rises less and less steeply to
maximum kappa = 0.35 approximately,
representing fair agreement between probablility of a true yes = 0.15
and 0.85.
The curve for 70% chance correct rises less and less steeply to
maximum kappa = 0.15 approximately, representing poor agreement throughout.
The curve for 60% chance correct rises less and less steeply to
maximum kappa = 0.05 approximately, representing poor agreement throughout.
Back to
Kappa statistics.
This page maintained by Martin Bland.
Last updated: 3 August, 2015.