Chi-square statistic for a categorical feature

Does this category tell us anything about the label?

A filter method scores each feature on its own, before any model is trained. For a categorical feature, the chi-square statistic compares how often each (category, label) pair actually occurs with how often it would occur if the feature and the label were independent.

Implement chi2_statistic(x, y).

• x and y are one-dimensional integer arrays of the same length (category and label).
• Build the contingency table O with one row per distinct category of x and one column per distinct label of y.
• Expected counts are E[i, j] = row_total[i] * col_total[j] / N.
• Return sum((O - E) ** 2 / E) as a Python float.

Constraints

• If x and y differ in length, or are empty, raise ValueError.
• Do not modify the inputs.

Hints