update

experiment_utils/experiment_analyzer.py

@@ -458,7 +458,7 @@ def get_effects(self, min_binary_count=100, adjustment=None):
 
         key_experiments = self.data.select(*self.experiment_identifier).distinct().collect()
 
-        results = []
+        temp_results = []
 
         if adjustment is None:
             adjustment = self.adjustment
@@ -486,7 +486,7 @@ def get_effects(self, min_binary_count=100, adjustment=None):
 
             treatvalues = set(temp_pd[self.treatment_col].unique())
             if len(treatvalues) != 2:
-                self.logger.warning('Skipping as it is not a valid treatment-control group!')
+                self.logger.warning('Skipping as there are no valid treatment-control groups!')
                 continue
             if not (0 in treatvalues and 1 in treatvalues):
                 log_and_raise_error(self.logger, f'The treatment column {self.treatment_col} must be 0 and 1')
@@ -517,10 +517,7 @@ def get_effects(self, min_binary_count=100, adjustment=None):
 
             if len(final_covariates) > 0:
                 temp_pd["weights"] = 1
-                temp_pd = self.standardize_covariates(
-                    temp_pd, final_covariates
-                )
-
+                temp_pd = self.standardize_covariates(temp_pd, final_covariates)
                 balance = self.calculate_smd(
                     data=temp_pd, covariates=final_covariates
                 )
@@ -529,7 +526,6 @@ def get_effects(self, min_binary_count=100, adjustment=None):
                 self._balance.append(balance)
                 self.logger.info('::::: Balance: %.2f', np.round(balance["balance_flag"].mean(), 2))
                 if adjustment == "IPW":
-                    temp_pd = self.standardize_covariates(temp_pd, final_covariates)
                     temp_pd = self.estimate_ipw(
                         data=temp_pd,
                         covariates=[f"z_{cov}" for cov in final_covariates],
@@ -561,11 +557,11 @@ def get_effects(self, min_binary_count=100, adjustment=None):
                 output['adjustment'] = 'No adjustment' if adjustment is None else adjustment
                 if adjustment == 'IPW':
                     output['balance'] = np.round(adjusted_balance['balance_flag'].mean(), 2)
-                elif len(final_covariates) > 0:
+                elif (len(final_covariates) > 0):
                     output['balance'] = np.round(balance['balance_flag'].mean(), 2)
                 output['experiment'] = experiment_tuple
 
-                results.append(output)
+                temp_results.append(output)
 
         result_columns = ['experiment', 'outcome', 'adjustment',
                           'treated_units', 'control_units', 'control_value',
@@ -577,15 +573,15 @@ def get_effects(self, min_binary_count=100, adjustment=None):
             index_to_insert = result_columns.index('adjustment') + 1
             result_columns.insert(index_to_insert, 'balance')
 
-        clean_results = pd.DataFrame(results)
-        clean_results = clean_results[result_columns]
+        clean_temp_results = pd.DataFrame(temp_results)
+        clean_temp_results = clean_temp_results[result_columns]
 
         if len(self._balance) > 0:
             self._balance = pd.concat(self._balance)
         if len(self._adjusted_balance) > 0:
             self._adjusted_balance = pd.concat(self._adjusted_balance)
 
-        self._results = self.__transform_tuple_column(clean_results, 'experiment', self.experiment_identifier)
+        self._results = self.__transform_tuple_column(clean_temp_results, 'experiment', self.experiment_identifier)
 
     def combine_effects(self, data: pd.DataFrame = None, grouping_cols: List = None):
         """
@@ -647,7 +643,7 @@ def __get_fixed_meta_analysis_estimate(self, data):
         except FloatingPointError:
             pvalue = np.nan
 
-        results = {
+        meta_results = {
             'experiments': int(data.shape[0]),
             'treated_units': int(data['treated_units'].sum()),
             'control_units': int(data['control_units'].sum()),
@@ -658,9 +654,9 @@ def __get_fixed_meta_analysis_estimate(self, data):
         }
 
         if 'balance' in data.columns:
-            results['balance'] = data['balance'].mean()
-        results['stat_significance'] = 1 if results['pvalue'] < self.alpha else 0
-        return results
+            meta_results['balance'] = data['balance'].mean()
+        meta_results['stat_significance'] = 1 if meta_results['pvalue'] < self.alpha else 0
+        return meta_results
 
     def aggregate_effects(self, data: pd.DataFrame = None, grouping_cols: List = None):
         """
@@ -689,17 +685,17 @@ def aggregate_effects(self, data: pd.DataFrame = None, grouping_cols: List = Non
             if 'outcome' not in grouping_cols:
                 grouping_cols.append('outcome')
 
-        results = data.groupby(grouping_cols).apply(self.__compute_weighted_effect).reset_index()
+        aggregate_results = data.groupby(grouping_cols).apply(self.__compute_weighted_effect).reset_index()
 
         self.logger.info('Aggregating effects using weighted averages!')
         self.logger.info('For a better standard error estimation, use meta-analysis or `combine_effects`')
 
         # keep initial order
         result_columns = grouping_cols + ['experiments', 'balance']
-        existing_columns = [col for col in result_columns if col in results.columns]
-        remaining_columns = [col for col in results.columns if col not in existing_columns]
+        existing_columns = [col for col in result_columns if col in aggregate_results.columns]
+        remaining_columns = [col for col in aggregate_results.columns if col not in existing_columns]
         final_columns = existing_columns + remaining_columns
-        return results[final_columns]
+        return aggregate_results[final_columns]
 
     def __compute_weighted_effect(self, group):