fixes for weighted kernel fits

statsmodels/nonparametric/kde.py

@@ -150,6 +150,7 @@ def fit(self, kernel="gau", bw="scott", fft=True, weights=None,
         self.kernel = kernel_switch[kernel](h=bw) # we instantiate twice,
                                                 # should this passed to funcs?
         # put here to ensure empty cache after re-fit with new options
+        self.kernel.weights = weights
         self._cache = resettable_cache()
 
     @cache_readonly
@@ -334,6 +335,8 @@ def kdensity(X, kernel="gau", bw="scott", weights=None, gridsize=None,
         weights = np.ones(nobs)
         q = nobs
     else:
+        # ensure weights is a numpy array
+        weights = np.asarray(weights)
         if len(weights) != len(clip_x):
             msg = "The length of the weights must be the same as the given X."
             raise ValueError(msg)

statsmodels/sandbox/nonparametric/kernels.py

@@ -48,6 +48,7 @@ def __init__(self, n, kernels = None, H = None):
             kernels = Gaussian()
 
         self._kernels = kernels
+        self.weights = None
 
         if H is None:
             H = np.matrix( np.identity(n))
@@ -66,13 +67,17 @@ def setH(self, value):
     H = property(getH, setH, doc="Kernel bandwidth matrix")
 
     def density(self, xs, x):
+
         n = len(xs)
         #xs = self.inDomain( xs, xs, x )[0]
 
         if len(xs)>0:  ## Need to do product of marginal distributions
             #w = np.sum([self(self._Hrootinv * (xx-x).T ) for xx in xs])/n
             #vectorized doesn't work:
-            w = np.mean(self((xs-x) * self._Hrootinv )) #transposed
+            if self.weights is not None:
+                w = np.mean(self((xs-x) * self._Hrootinv).T * self.weights)/sum(self.weights)
+            else:
+                w = np.mean(self((xs-x) * self._Hrootinv )) #transposed
             #w = np.mean([self(xd) for xd in ((xs-x) * self._Hrootinv)] ) #transposed
             return w
         else:
@@ -137,6 +142,7 @@ def __init__(self, shape, h = 1.0, domain = None, norm = None):
             norm = 1.0
         self._normconst = norm
         self.domain = domain
+        self.weights = None
         if callable(shape):
             self._shape = shape
         else:
@@ -185,7 +191,10 @@ def density(self, xs, x):
             xs = xs[:,None]
         if len(xs)>0:
             h = self.h
-            w = 1/h * np.mean(self((xs-x)/h), axis=0)
+            if self.weights is not None:
+                w = 1/h * np.sum(self((xs-x)/h).T * self.weights, axis=1)
+            else:
+                w = 1/h * np.mean(self((xs-x)/h), axis=0)
             return w
         else:
             return np.nan