dualbounds.dist_reg.DistReg.predict_counterfactuals¶

DistReg.predict_counterfactuals(X: array)[source]¶

Predicts counterfactual distributions of Y (outcome).

Parameters:¶

X : np.array¶

(n,p)-shaped design matrix.

Returns:¶

• y0_dists (np.array) – y0_dists[i] = the law of \(Y_i(0) | X_i\). Only returned in a setting without instrumental variables.

• y1_dists (np.array) – y1_dists[i] = the law of \(Y_i(1) | X_i\). Only returned in a setting without instrumental variables.

• ydists (list) – Only returned if trained using instrumental variables. Then ydists[z][w] is a batched scipy distribution array whose ith element represents the law of \(Y_i(w) | X_i, W_i(z) = w\).