Categorical Correlation

A simple wrapper for some useful correlation/association metrics for categorical/nominal variables.

Available metrics:

• Cramer's V
• Tschuprow's T
• Pearson Contingency Value
• Theil's U (assymetric)

Example

Let's use some random categorical data as an example:

import pandas as pd
import numpy as np

data = pd.util.testing.makeCustomDataframe(
    nrows=10000,
    ncols=10,
    dtype='category',
    data_gen_f=lambda *args: np.random.randint(5)
)

Then, let's generate the cramer correlation matrix.

import cat_corr as cat
res = cat.get_categorical_corr(
    data=data,
    features=data.columns,
    method='cramer',
    thr=0.5
)

cramer_corr_matrix = res[0]

Requirements

• pandas >= 1.0.0
• numpy >= 1.21.2
• scipy >= 1.7.1
• dython >= 0.7.1.post3