Method Selection GuideΒΆ

The following table shows which correlation method to use based on your variable types:

Variable X

Variable Y

Method

dichotomous (discretized)

dichotomous (discretized)

ordinalcorr.tetrachoric()

polytomous (discretized)

polytomous (discretized)

ordinalcorr.polychoric()

continuous

polytomous (discretized)

ordinalcorr.polyserial()

continuous

dichotomous (discretized)

ordinalcorr.biserial()

continuous

dichotomous

ordinalcorr.point_biserial()

Where:

  • dichotomous variable: An ordinal variable with exactly two categories (e.g., Yes/No, 0/1).

  • polytomous variable: An ordinal variable with more than two categories (e.g., Likert scale with 5 options).

  • discretized: Indicates that the variable is assumed to originate from an underlying continuous latent distribution, and that observed categories result from applying thresholds to this latent variable.