update to support more scikit models

.travis.yml

@@ -10,41 +10,56 @@ python:
   - 3.6
   - 3.7
   - 3.8
+  - "3.10"
+  - "3.11"
 env:
   - SKLEARNVERSION=0.23.2
   - SKLEARNVERSION=0.24
-matrix:
-  include:      
-    - name: "Python 3.7 on Windows sk-learn v0.23"
-      os: windows           # Windows 10.0.17134 N/A Build 17134
-      language: shell
-      before_install:
-        - choco install python --version 3.7
-        - python -m pip install --upgrade pip
-        - pip install nose
-      env: 
-       - PATH=/c/Python37:/c/Python37/Scripts:$PATH
-       - SKLEARNVERSION=0.23.2
-    - name: "Python 3.7 on Windows sk-learn v0.24"
-      os: windows           # Windows 10.0.17134 N/A Build 17134
-      language: shell
-      before_install:
-        - choco install python --version 3.7
-        - python -m pip install --upgrade pip
-        - pip install nose
-      env: 
-       - PATH=/c/Python37:/c/Python37/Scripts:$PATH
-       - SKLEARNVERSION=0.24
-    - name: "Python 3.8 on Windows sk-learn v0.24"
-      os: windows           # Windows 10.0.17134 N/A Build 17134
-      language: shell
-      before_install:
-        - choco install python --version 3.8
-        - python -m pip install --upgrade pip
-        - pip install nose
-      env: 
-       - PATH=/c/Python38:/c/Python38/Scripts:$PATH
-       - SKLEARNVERSION=0.24
+  - SKLEARNVERSION=0.24.1
+  - SKLEARNVERSION=0.24.2
+  - SKLEARNVERSION=1.0
+  - SKLEARNVERSION=1.0.1
+  - SKLEARNVERSION=1.0.2
+  - SKLEARNVERSION=1.1.0
+  - SKLEARNVERSION=1.1.1
+  - SKLEARNVERSION=1.1.2
+  - SKLEARNVERSION=1.1.3
+  - SKLEARNVERSION=1.2.0
+  - SKLEARNVERSION=1.2.1
+  - SKLEARNVERSION=1.2.2
+  - SKLEARNVERSION=1.3.0
+#matrix:
+#  include:      
+#    - name: "Python 3.7 on Windows sk-learn v0.23"
+#      os: windows           # Windows 10.0.17134 N/A Build 17134
+#      language: shell
+#      before_install:
+#        - choco install python --version 3.7
+#        - python -m pip install --upgrade pip
+#        - pip install nose
+#      env: 
+#       - PATH=/c/Python37:/c/Python37/Scripts:$PATH
+#       - SKLEARNVERSION=0.23.2
+#    - name: "Python 3.7 on Windows sk-learn v0.24"
+#      os: windows           # Windows 10.0.17134 N/A Build 17134
+#      language: shell
+#      before_install:
+#        - choco install python --version 3.7
+#        - python -m pip install --upgrade pip
+#        - pip install nose
+#      env: 
+#       - PATH=/c/Python37:/c/Python37/Scripts:$PATH
+#       - SKLEARNVERSION=0.24
+#    - name: "Python 3.8 on Windows sk-learn v0.24"
+#      os: windows           # Windows 10.0.17134 N/A Build 17134
+#      language: shell
+#      before_install:
+#        - choco install python --version 3.8
+#        - python -m pip install --upgrade pip
+#        - pip install nose
+#      env: 
+#       - PATH=/c/Python38:/c/Python38/Scripts:$PATH
+#       - SKLEARNVERSION=0.24
 before_install:
  - pip3 install cpp-coveralls
 install:

dame_flame/data_cleaning.py

@@ -119,11 +119,14 @@ def check_parameters(adaptive_weights, df_holdout, df_input, alpha, FLAME,
 
         # make sure that adaptive_weights is a valid value.
         if (adaptive_weights not in ["ridge", "decisiontree", "ridgeCV", "decisiontreeCV"]):
-            raise Exception("Invalid input error. The acceptable values for "\
+            # Check to see if adaptive_weights is an object of type scikit-model
+            if not (hasattr(adaptive_weights, 'fit') and hasattr(adaptive_weights, 'predict')):
+                raise Exception("Invalid input error. The acceptable values for "\
                             "the adaptive_weights parameter are 'ridge', "\
                             "'decisiontree', 'decisiontreeCV', or 'ridgeCV'. Additionally, "\
                             "adaptive-weights may be 'False' along "\
-                            "with a weight array")
+                            "with a weight array or may be a scikit learn object if it has"\
+                            " a fit, predict method.")
 
 
         # make sure the two dfs have the same number of columns first:

dame_flame/flame_dame_helpers.py

@@ -96,11 +96,21 @@ def find_pe_for_covar_set(df_holdout, treatment_column_name,
             clf = Ridge(alpha=alpha_given)
         elif adaptive_weights in ["decisiontree", "decisiontreeCV"]:
             clf = DecisionTreeRegressor()
-        else:
+        elif adaptive_weights == 0:
             return False
+        else:
+            clf = adaptive_weights
+            # return False
 
-        if adaptive_weights in ["ridge", "decisiontree"]:
-
+        if adaptive_weights in ["ridgeCV", "decisiontreeCV"]:
+            # calculate MSE via cross validation.
+            mse_treated = -1*np.mean(cross_val_score(clf, x_treated, y_treated,
+                                                     scoring='neg_mean_squared_error',
+                                                     cv=5))
+            mse_control = -1*np.mean(cross_val_score(clf, x_control, y_control,
+                                                     scoring='neg_mean_squared_error',
+                                                     cv=5))
+        else:
             # Calculate treated MSE
             clf.fit(x_treated, y_treated)
             predicted = clf.predict(x_treated)
@@ -111,15 +121,6 @@ def find_pe_for_covar_set(df_holdout, treatment_column_name,
             predicted = clf.predict(x_control)
             mse_control = mean_squared_error(y_control, predicted)
 
-        else:
-            # calculate MSE via cross validation.
-            mse_treated = -1*np.mean(cross_val_score(clf, x_treated, y_treated,
-                                                     scoring='neg_mean_squared_error',
-                                                     cv=5))
-            mse_control = -1*np.mean(cross_val_score(clf, x_control, y_control,
-                                                     scoring='neg_mean_squared_error',
-                                                     cv=5))
-
         pe_array.append(mse_treated + mse_control)
 
     return np.mean(pe_array)

dame_flame/matching.py

@@ -7,6 +7,7 @@
 # License: MIT
 
 import numpy as np
+import pandas as pd
 from . import data_cleaning
 from . import dame_algorithm
 from . import flame_algorithm
@@ -242,7 +243,9 @@ def predict(self, input_data, pre_dame=float('inf'), C=0.1):
 
             # the first few items all look the same, then the last item is from dame
             self.df_units_and_covars_matched = return_array[0]
-            self.df_units_and_covars_matched = self.df_units_and_covars_matched.append(return_array[-1][0], sort=True)
+            # self.df_units_and_covars_matched = self.df_units_and_covars_matched.append(return_array[-1][0], sort=True)
+            self.df_units_and_covars_matched = pd.concat([self.df_units_and_covars_matched, return_array[-1][0]])
+
             if self.repeats == True:
                 # we have to aggregate the indexes appearing more than once,
                 # those are the ones which were matched units in both dame and flame: