ordinalcorr: correlation coefficients for ordinal variables

ordinalcorr is a Python package designed to compute correlation coefficients tailored for ordinal-scale data (e.g., Likert items). It supports polychoric correlation coefficients and other coefficients for ordinal data.

✨ Features

1️⃣ Correlation Coefficients

This package provides several correlation coefficients (e.g. Polyserial and Polychoric)

Here is an example:

>>> from ordinalcorr import polychoric
>>> x = [1, 1, 2, 2, 3, 3]
>>> y = [0, 0, 0, 1, 1, 1]
>>> polychoric(x, y)
0.9986287922233864

2️⃣ Heterogeneous Correlation Matrix

A function for computing the heterogeneous correlation matrix—a correlation matrix that includes both continuous and ordinal variables—is also provided.

>>> from ordinalcorr import hetcor
>>> import pandas as pd
>>> import numpy as np
>>> np.random.seed(0)
>>> data = pd.DataFrame({
...     "continuous": np.random.normal(size=90),
...     "dichotomous": np.repeat([0, 1], 45),
...     "polytomous": np.repeat([1, 2, 3], 30),
... })
>>> hetcor(data).round(3)
             continuous  dichotomous  polytomous
continuous        1.000       -0.254      -0.257
dichotomous      -0.254        1.000       0.999
polytomous       -0.257        0.999       1.000

📦 Installation

ordinalcorr is available at the PyPI

pip install ordinalcorr

Requirements

• Python 3.10 or later
• Dependencies:
  • numpy >= 1.23.0
  • scipy >= 1.8.0

📒 Documentation

ordinalcorr documentation

⚖️ License

This project is licensed under the MIT License. See the LICENSE file for details.