Analyses

Goodman–Kruskal lambda

How much knowing one category reduces errors about another.

Edit on GitHub

When to use it

Use lambda when both variables are categories, ordered or not, and you want the reduction in prediction errors. Lambda is 0 when the extra variable does not help, and 1 when it predicts the other variable perfectly.

Assumptions

Both columns are categorical. Lambda can be zero even when chi-square is significant, because it asks a stricter prediction question.

Running it in Tensr

Analyze → Correlate → Lambda. In chat: “Lambda for dose and response.”
Column A and column B are the two categorical variables.

Options

Prop

Type

Reading the output

How much knowing the teaching method reduces errors in predicting the satisfaction rating.

method1234
Lecture512123
Online41279
Workshop47138

Goodman–Kruskal λ for method and satisfaction, n = 96. The primary result is significant (p = < .001). This is large enough, in this sample, to treat the comparison this page is about as a real association rather than noise. Metrics: λ (symmetric) = 0.301; λ (A|B) = 0.289; λ (B|A) = 0.313; ASE = 0.047; z = 6.433; p-value = < .001.

Reporting (APA 7)

Goodman–Kruskal λ for method and satisfaction, n = 96. This result is significant (p = < .001). The effect is moderate, so report its size with the p-value.

Coming from SPSS

Analyze → Correlate → Lambda.