Fixed proposal functions for Metropolis

Author: Christopher Fonnesbeck

pymc/step_methods/hmc.py

@@ -84,5 +84,5 @@ def astep(self, q0, logp, dlogp):
 
         self.state.metrops.append(mr)
 
-        return metrop_select(mr, q, q0)[0]
+        return metrop_select(mr, q, q0)
 

pymc/step_methods/metropolis.py

@@ -10,13 +10,19 @@
 
 # Available proposal distributions for Metropolis
 def normal_proposal(s):
-    return lambda: normal(scale=s)
+    def random():
+        return normal(scale=s)
+    return random
 
 def cauchy_proposal(s):
-    return lambda: standard_cauchy(size=np.size(s)) * s
+    def random():
+        return standard_cauchy(size=np.size(s)) * s
+    return random
 
 def laplace_proposal(s):
-    return lambda: (standard_exponential(size=np.size(s)) - standard_exponential(size=np.size(s))) * s
+    def random():
+        return (standard_exponential(size=np.size(s)) - standard_exponential(size=np.size(s))) * s
+    return random
 
 
 class Metropolis(ArrayStep):