Merge pull request #427 from UCL-CCS/feature_more_stats_in_qmc

more stats for (q)mc

easyvvuq/analysis/qmc_analysis.py

@@ -51,17 +51,21 @@ def supported_stats(self):
         -------
         list of str
         """
-        return ['mean', 'var', 'std', 'percentiles', '10%', '50%', '90%']
+        return ['mean', 'var', 'std', 'min', 'max', 'median', 'percentiles', '1%', '10%', '50%', '90%', '99%']
 
     def _describe(self, qoi, statistic):
         if statistic not in self.supported_stats():
             raise NotImplementedError
+        if statistic == '1%':
+            return self.raw_data['percentiles'][qoi]['p1']
         if statistic == '10%':
             return self.raw_data['percentiles'][qoi]['p10']
         elif statistic == '50%':
             return self.raw_data['percentiles'][qoi]['p50']
         elif statistic == '90%':
             return self.raw_data['percentiles'][qoi]['p90']
+        elif statistic == '99%':
+            return self.raw_data['percentiles'][qoi]['p99']
         else:
             return self.raw_data['statistical_moments'][qoi][statistic][0]
 
@@ -206,10 +210,16 @@ def analyse(self, data_frame):
 
             results['statistical_moments'][k] = {'mean': np.mean(masked_samples, axis=0),
                                                  'var': np.var(masked_samples, axis=0),
-                                                 'std': np.std(masked_samples, axis=0)}
-            results['percentiles'][k] = {'p10': np.percentile(masked_samples, 10, 0)[0],
+                                                 'std': np.std(masked_samples, axis=0),
+                                                 'min': np.min(masked_samples, axis=0),
+                                                 'max': np.max(masked_samples, axis=0),
+                                                 'median': np.median(masked_samples, axis=0),
+                                                 }
+            results['percentiles'][k] = {'p1': np.percentile(masked_samples, 1, 0)[0],
+                                         'p10': np.percentile(masked_samples, 10, 0)[0],
                                          'p50': np.percentile(masked_samples, 50, 0)[0],
-                                         'p90': np.percentile(masked_samples, 90, 0)[0]}                                                 
+                                         'p90': np.percentile(masked_samples, 90, 0)[0],
+                                         'p99': np.percentile(masked_samples, 99, 0)[0]}
 
             # Replace Nan values by the mean before proceeding with the SA
             indices = np.where(mask == 0)[0] # samples[~mask] = results[k].mean

tests/test_mc_analysis_results.py

@@ -119,18 +119,28 @@ def test_describe(results_vectors):
         results_vectors.describe()[
             ('g',
              1)].to_dict() == {
+            '1%': 0.007945053328978277,
             '10%': 0.08156520178597204,
-            '90%': 0.8729378821725343,     
+            '90%': 0.8729378821725343,
+            '99%': 0.9950881374303691,
+            'max': 0.9983981908285983,
             'mean': 0.4691844466934421,
-            'var': 0.08534945020531205,
-            'std': 0.29214628220347433})
+            'median': 0.4495551186167759,
+            'min': 0.0038532031638378594,
+            'std': 0.29214628220347433,
+            'var': 0.08534945020531205})
     assert (
         results_vectors.describe('h')[
             ('h',
              1)].to_dict() == {
+            '1%': 0.039851582820686876,
             '10%': 0.21724677702965456,
             '90%': 0.9764815719704141,
+            '99%': 0.9969701901368494,
+            'max': 0.9993165378092128,
             'mean': 0.6873389710989142,
-            'var': 0.07501266456861228,
-            'std': 0.27388440000958847})
+            'median': 0.7696583654805108,
+            'min': 0.01590577381510083,
+            'std': 0.27388440000958847,
+            'var': 0.07501266456861228})
     assert (isinstance(results_vectors.describe('h', 'std'), np.ndarray))