Fix scoring kwargs call (#33)

Fix a bug when calling `OutcomeEvaluator` with custom metrics due to
positional arguments misalignment.

Fix an issue with Travis failing to install in a Python 3.7 build.
* https://app.travis-ci.com/github/IBM/causallib/jobs/565683172
* pytest-dev/pytest#7371 (comment)
* tobinus/python-podgen#124


Co-authored-by: ehudkr <ehudkaravani@gmail.com>

.travis.yml

@@ -11,6 +11,7 @@ before_script:
   - ./cc-test-reporter before-build
 install:
   - pip install --upgrade pip
+  - pip install --upgrade importlib-metadata  # Solve a python 3.7 install bug: https://app.travis-ci.com/github/IBM/causallib/jobs/566048347
   - pip install -r requirements.txt
   - pip install -r causallib/contrib/requirements.txt
   - pip install --upgrade pytest coverage

causallib/evaluation/outcome_evaluator.py

@@ -139,9 +139,18 @@ def score_estimation(self, prediction, X, a_true, y_true, metrics_to_evaluate=No
     def _score_single(self, y_true, prediction, prediction_prob, outcome_is_binary, metrics_to_evaluate):
         """Score a single prediction based on whether `y_true` is classification or regression"""
         if outcome_is_binary:
-            score = self.score_binary_prediction(y_true, prediction, prediction_prob, metrics_to_evaluate)
+            score = self.score_binary_prediction(
+                y_true=y_true,
+                y_pred=prediction,
+                y_pred_proba=prediction_prob,
+                metrics_to_evaluate=metrics_to_evaluate
+            )
         else:
-            score = self.score_regression_prediction(y_true, prediction, metrics_to_evaluate)
+            score = self.score_regression_prediction(
+                y_true=y_true,
+                y_pred=prediction,
+                metrics_to_evaluate=metrics_to_evaluate
+            )
         # score = pd.DataFrame(score).T
         # score = score.apply(pd.to_numeric, errors="ignore")  # change dtype of each column to numerical if possible.
         return score