nispace.stats.coloc.mutualinfo

nispace.stats.coloc.mutualinfo(x, y, n_neighbors=3, seed=None)[source]

Compute mutual information between x and y via sklearn’s k-NN estimator.

Thin wrapper around sklearn.feature_selection.mutual_info_regression (not numba-jitted).

Parameters:

  • x (np.ndarray, shape (n_obs,) or (n_obs, 1)) – 1D input is reshaped to a column vector.

  • y (np.ndarray, shape (n_obs,)) –

  • n_neighbors (int, default 3) – Number of neighbors for the k-NN MI estimator.

  • seed (int, optional) – Passed as mutual_info_regression’s random_state. The estimator adds small random noise to break ties, so results are not reproducible across calls unless this is set.

Returns:

mi

Return type:

float

Notes

Not NaN-tolerant (sklearn errors on NaN) – called with pre-masked, NaN-free x[mask]/y[mask] in core/colocalize.py’s “mi” colocalization method.