Merge pull request #147 from kyleabeauchamp/minor

Added W_nk property and updated a test for n_eff_samples

pymbar/mbar.py

@@ -274,6 +274,20 @@ def __init__(self, u_kn, N_k, maximum_iterations=10000, relative_tolerance=1.0e-
             print "MBAR initialization complete."
         return
 
+    @property
+    def W_nk(self):
+        """Retrieve the weight matrix W_nk from the MBAR algorithm.
+
+        Necessary because they are stored internally as log weights.
+
+        Returns
+        -------
+        weights : np.ndarray, float, shape=(N, K)
+            NxK matrix of weights in the MBAR covariance and averaging formulas
+
+        """        
+        return np.exp(self.Log_W_nk)
+
 
     #=========================================================================
     def getWeights(self):
@@ -288,7 +302,7 @@ def getWeights(self):
 
         """
 
-        return np.exp(self.Log_W_nk)
+        return self.W_nk
 
     #=========================================================================
     def computeEffectiveSampleNumber(self, verbose = False):
@@ -340,7 +354,6 @@ def computeEffectiveSampleNumber(self, verbose = False):
         for k in range(self.K):
             w = np.exp(self.Log_W_nk[:,k])
             N_eff[k] = 1/np.sum(w**2)
-
             if verbose:
                 print "Effective number of sample in state %d is %10.3f" % (k,N_eff[k])
                 print "Efficiency for state %d is %d/%d = %10.4f" % (k,N_eff[k],len(w),N_eff[k]/len(w))

pymbar/tests/test_mbar.py

@@ -199,8 +199,8 @@ def test_mbar_computeEffectiveSampleNumber():
         # one mathematical effective sample numbers should be between N_k and sum_k N_k
         N_eff = mbar.computeEffectiveSampleNumber()
         sumN = np.sum(N_k)
-        for k in range(len(N_eff)):
-            eq(np.bool(N_eff[k] > N_k[k] and N_eff[k] < sumN),True)
+        assert all(N_eff > N_k)
+        assert all(N_eff < sumN)
 
 def test_mbar_computeOverlap():