cohen_kappa#

scikitplot.corpus.cohen_kappa(rater_a, rater_b, *, labels=None, weights=None)[source]#

Cohen’s kappa between two raters of the same items.

Parameters:
rater_a, rater_bsequence

One label per item, in the same item order; no missing values.

labelssequence or None, default=None

Label order (it matters for weights); None sorts the labels seen.

weights{None, “linear”, “quadratic”}, default=None

Disagreement weights: none (each disagreement counts 1), |i - j| or (i - j) ** 2 over label positions.

Returns:
float

NaN (with a warning) when chance agreement is already perfect.

Raises:
ValueError

For different lengths, no items, missing values or an unknown weight.

Parameters:

  • rater_a (Any)

  • rater_b (Any)

  • labels (Any)

  • weights (str | None)

Return type:

float

Examples

>>> round(cohen_kappa(["a", "b", "a"], ["a", "b", "b"]), 4)
0.4