From c891f465c921e31bbb1de3ea9c71437aa9d4b7c6 Mon Sep 17 00:00:00 2001
From: Jeong-Yoon Lee <jeongyoon.lee1@gmail.com>
Date: Thu, 9 May 2024 11:25:09 -0700
Subject: [PATCH] fix #771 and add a test for get/plot_tmlegain()

---
 causalml/metrics/visualize.py | 20 ++++++-----
 pyproject.toml                |  2 +-
 tests/test_visualize.py       | 64 +++++++++++++++++++++++++++++++++--
 3 files changed, 75 insertions(+), 11 deletions(-)

diff --git a/causalml/metrics/visualize.py b/causalml/metrics/visualize.py
index 39fe4bf7..7dd13980 100644
--- a/causalml/metrics/visualize.py
+++ b/causalml/metrics/visualize.py
@@ -340,12 +340,16 @@ def get_tmlegain(
     lift_ub = []
 
     for col in model_names:
+        # Create `n_segment` equal segments from sorted model estimates. Rank is used to break ties.
+        # ref: https://stackoverflow.com/a/46979206/3216742
+        segments = pd.qcut(df[col].rank(method="first"), n_segment, labels=False)
+
         ate_model, ate_model_lb, ate_model_ub = tmle.estimate_ate(
             X=df[inference_col],
             p=df[p_col],
             treatment=df[treatment_col],
             y=df[outcome_col],
-            segment=pd.qcut(df[col], n_segment, labels=False),
+            segment=segments,
         )
         lift_model = [0.0] * (n_segment + 1)
         lift_model[n_segment] = ate_all[0]
@@ -446,19 +450,21 @@ def get_tmleqini(
     qini_ub = []
 
     for col in model_names:
+        # Create `n_segment` equal segments from sorted model estimates. Rank is used to break ties.
+        # ref: https://stackoverflow.com/a/46979206/3216742
+        segments = pd.qcut(df[col].rank(method="first"), n_segment, labels=False)
+
         ate_model, ate_model_lb, ate_model_ub = tmle.estimate_ate(
             X=df[inference_col],
             p=df[p_col],
             treatment=df[treatment_col],
             y=df[outcome_col],
-            segment=pd.qcut(df[col], n_segment, labels=False),
+            segment=segments,
         )
 
         qini_model = [0]
         for i in range(1, n_segment):
-            n_tr = df[pd.qcut(df[col], n_segment, labels=False) == (n_segment - i)][
-                treatment_col
-            ].sum()
+            n_tr = df[segments == (n_segment - i)][treatment_col].sum()
             qini_model.append(ate_model[0][n_segment - i] * n_tr)
 
         qini.append(qini_model)
@@ -467,9 +473,7 @@ def get_tmleqini(
             qini_lb_model = [0]
             qini_ub_model = [0]
             for i in range(1, n_segment):
-                n_tr = df[pd.qcut(df[col], n_segment, labels=False) == (n_segment - i)][
-                    treatment_col
-                ].sum()
+                n_tr = df[segments == (n_segment - i)][treatment_col].sum()
                 qini_lb_model.append(ate_model_lb[0][n_segment - i] * n_tr)
                 qini_ub_model.append(ate_model_ub[0][n_segment - i] * n_tr)
 
diff --git a/pyproject.toml b/pyproject.toml
index 4c599fd3..286c9c1b 100755
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "causalml"
-version = "0.15.1"
+version = "0.15.2dev"
 description = "Python Package for Uplift Modeling and Causal Inference with Machine Learning Algorithms"
 readme = { file = "README.md", content-type = "text/markdown" }
 
diff --git a/tests/test_visualize.py b/tests/test_visualize.py
index f3bc0669..414247a0 100644
--- a/tests/test_visualize.py
+++ b/tests/test_visualize.py
@@ -1,7 +1,11 @@
-import pandas as pd
+from matplotlib import pyplot as plt
 import numpy as np
+import pandas as pd
 import pytest
-from causalml.metrics.visualize import get_cumlift
+from sklearn.model_selection import KFold, train_test_split
+
+from causalml.metrics.visualize import get_cumlift, plot_tmlegain
+from causalml.inference.meta import LRSRegressor
 
 
 def test_visualize_get_cumlift_errors_on_nan():
@@ -12,3 +16,59 @@ def test_visualize_get_cumlift_errors_on_nan():
 
     with pytest.raises(Exception):
         get_cumlift(df)
+
+
+def test_plot_tmlegain(generate_regression_data, monkeypatch):
+    monkeypatch.setattr(plt, "show", lambda: None)
+
+    y, X, treatment, tau, b, e = generate_regression_data()
+
+    (
+        X_train,
+        X_test,
+        y_train,
+        y_test,
+        e_train,
+        e_test,
+        treatment_train,
+        treatment_test,
+        tau_train,
+        tau_test,
+        b_train,
+        b_test,
+    ) = train_test_split(X, y, e, treatment, tau, b, test_size=0.5, random_state=42)
+
+    learner = LRSRegressor()
+    learner.fit(X_train, treatment_train, y_train)
+    cate_test = learner.predict(X_test, treatment_test).flatten()
+
+    df = pd.DataFrame(
+        {
+            "y": y_test,
+            "w": treatment_test,
+            "p": e_test,
+            "S-Learner": cate_test,
+            "Actual": tau_test,
+        }
+    )
+
+    inference_cols = []
+    for i in range(X_test.shape[1]):
+        col = "col_" + str(i)
+        df[col] = X_test[:, i]
+        inference_cols.append(col)
+
+    n_fold = 3
+    kf = KFold(n_splits=n_fold)
+
+    plot_tmlegain(
+        df,
+        inference_col=inference_cols,
+        outcome_col="y",
+        treatment_col="w",
+        p_col="p",
+        n_segment=5,
+        cv=kf,
+        calibrate_propensity=True,
+        ci=False,
+    )