Merge branch 'feature/non_sample_size_tests' into develop

benford/benford.py

@@ -14,7 +14,7 @@
 from .reports import _inform_, _report_mad_, _report_test_, _deprecate_inform_,\
     _report_mantissa_
 from .stats import Z_score, chi_sq, chi_sq_2, kolmogorov_smirnov,\
-    kolmogorov_smirnov_2
+    kolmogorov_smirnov_2, bhattacharyya_distance, kullback_leibler_divergence
 
 
 class Base(DataFrame):
@@ -129,6 +129,10 @@ def __init__(self, base, digs, confidence, limit_N=None, sec_order=False):
         self.KS = kolmogorov_smirnov_2(self)
         self.MAD = self.AbsDif.mean()
         self.MSE = (self.AbsDif ** 2).mean()
+        self.bhattacharyya_distance = bhattacharyya_distance(
+            self.Found.values, self.Expected.values)
+        self.kullback_leibler_divergence = kullback_leibler_divergence(
+            self.Found.values, self.Expected.values)
         self.confidence = confidence
         self.digs = digs
         self.sec_order = sec_order

benford/reports.py

@@ -100,6 +100,18 @@ def _report_summ_(test, high_diff):
         print(test.sort_values('AbsDif', ascending=False))
 
 
+def _report_bhattac_dist_(bhattac_dist):
+    """
+    """
+    print(f"The Bhattacharyya Distance is: {bhattac_dist:6f}\n")
+
+
+def _report_kl_diverg_(kl_diverg):
+    """
+    """
+    print(f"The Kullback-Leibler Divergence is: {kl_diverg:6f}\n")
+
+
 def _report_test_(test, high=None, crit_vals=None):
     """Main report function. Receives the Args: to report with, initiates
     the process, and calls the right reporting helper function(s), depending
@@ -108,6 +120,8 @@ def _report_test_(test, high=None, crit_vals=None):
     print('\n', f'  {test.name}  '.center(50, '#'), '\n')
     if not 'Summation' in test.name:
         _report_mad_(test.digs, test.MAD)
+        _report_bhattac_dist_(test.bhattacharyya_distance)
+        _report_kl_diverg_(test.kullback_leibler_divergence)
         if test.confidence is not None:
             print(f"For confidence level {test.confidence}%: ")
             _report_KS_(test.KS, crit_vals['KS'])

benford/stats.py

@@ -1,4 +1,4 @@
-from numpy import sqrt
+from numpy import sqrt, log, where
 from .constants import crit_chi2, KS_crit, mad_dict, digs_dict
 
 
@@ -159,3 +159,50 @@ def mse(frame, verbose=True):
         print(f"\nMean Square Error = {mse}")
 
     return mse
+
+def _bhattacharyya_coefficient(dist_1, dist_2):
+    """Computes the Bhattacharyya Coeficient between two probability
+    distributions, to be letar used to compute the Bhattacharyya Distance
+
+    Args:
+        dist_1 (np.array): The newly gathered distribution, to be compared
+            with an older / established distribution.
+        dist_2 (np.array): The older/ establhished distribution with which
+            the new one will be compared. 
+    
+    Returns:
+        bhat_coef (float)
+    """
+    return sqrt(dist_1 * dist_2).sum()
+
+
+def bhattacharyya_distance(dist_1, dist_2):
+    """Computes the Bhattacharyya Dsitance between two probability
+    distributions
+
+    Args:
+        dist_1 (np.array): The newly gathered distribution, to be compared
+            with an older / established distribution.
+        dist_2 (np.array): The older/ establhished distribution with which
+            the new one will be compared. 
+    
+    Returns:
+        bhat_dist (float)
+    """
+    return -log(_bhattacharyya_coefficient(dist_1, dist_2))
+
+
+def kullback_leibler_divergence(dist_1, dist_2):
+    """Computes the Kullback-Leibler Divergence between two probability
+    distributions.
+
+    Args:
+        dist_1 (np.array): The newly gathered distribution, to be compared
+            with an older / established distribution.
+        dist_2 (np.array): The older/ establhished distribution with which
+            the new one will be compared. 
+
+    Returns:
+        kulb_leib_diverg (float)        
+    """
+    return (log((dist_1 / dist_2), where=(dist_1 != 0)) * dist_1).sum()