diff --git a/src/pg/sql/29_gwr_predict.sql b/src/pg/sql/29_gwr_predict.sql
new file mode 100644
index 0000000..de5fb4d
--- /dev/null
+++ b/src/pg/sql/29_gwr_predict.sql
@@ -0,0 +1,11 @@
+CREATE OR REPLACE FUNCTION
+CDB_GWR_PREDICT(subquery text, dep_var text, ind_vars text[],
+       bw numeric default null, fixed boolean default False, kernel text default 'bisquare')
+RETURNS table(coeffs JSON, stand_errs JSON, t_vals JSON, r_squared numeric, predicted numeric, rowid bigint)
+AS $$
+
+from crankshaft.regression import gwr_cs
+
+return gwr_cs.gwr_predict(subquery, dep_var, ind_vars, bw, fixed, kernel)
+
+$$ LANGUAGE plpythonu;
diff --git a/src/py/crankshaft/crankshaft/pysal_utils/pysal_utils.py b/src/py/crankshaft/crankshaft/pysal_utils/pysal_utils.py
index 4fbd3d6..0739fcb 100644
--- a/src/py/crankshaft/crankshaft/pysal_utils/pysal_utils.py
+++ b/src/py/crankshaft/crankshaft/pysal_utils/pysal_utils.py
@@ -215,6 +215,29 @@ def gwr_query(params):
 
     return query.format(**params).strip()
 
+def gwr_predict_query(params):
+    """
+    GWR query
+    """
+
+    replacements = {"ind_vars_select": query_attr_select(params,
+                                                         table_ref=None),
+                    "ind_vars_where": query_attr_where(params,
+                                                       table_ref=None)}
+
+    query = '''
+      SELECT
+        array_agg(ST_X(ST_Centroid({geom_col}))) As x,
+        array_agg(ST_Y(ST_Centroid({geom_col}))) As y,
+        array_agg({dep_var}) As dep_var,
+        %(ind_vars_select)s
+        array_agg({id_col}) As rowid
+      FROM ({subquery}) As q
+      WHERE
+        %(ind_vars_where)s
+        ''' % replacements
+
+    return query.format(**params).strip()
 # to add more weight methods open a ticket or pull request
 
 
diff --git a/src/py/crankshaft/crankshaft/regression/gwr_cs.py b/src/py/crankshaft/crankshaft/regression/gwr_cs.py
index ded9512..a1fe9c5 100644
--- a/src/py/crankshaft/crankshaft/regression/gwr_cs.py
+++ b/src/py/crankshaft/crankshaft/regression/gwr_cs.py
@@ -90,3 +90,100 @@ def gwr(subquery, dep_var, ind_vars, bw=None,
                         predicted, residuals, r_squared, rowid, bw)))
     return zip(coefficients, stand_errs, t_vals,
                predicted, residuals, r_squared, rowid, bw)
+
+def gwr_predict(subquery, dep_var, ind_vars, bw=None,
+        fixed=False, kernel='bisquare'):
+    """
+    subquery: 'select * from demographics'
+    dep_var: 'pctbachelor'
+    ind_vars: ['intercept', 'pctpov', 'pctrural', 'pctblack']
+    bw: value of bandwidth, if None then select optimal
+    fixed: False (kNN) or True ('distance')
+    kernel: 'bisquare' (default), or 'exponential', 'gaussian'
+    """
+
+    # query_result = subquery
+    params = {'geom_col': 'the_geom',
+              'id_col': 'cartodb_id',
+              'subquery': subquery,
+              'dep_var': dep_var,
+              'ind_vars': ind_vars}
+
+    try:
+        query = pu.gwr_predict_query(params)
+        plpy.notice(query)
+        query_result = plpy.execute(query)
+    except plpy.SPIError, err:
+        plpy.notice(query)
+        plpy.error('Analysis failed: %s' % err)
+
+    # unique ids and variable names list
+    rowid = np.array(query_result[0]['rowid'], dtype=np.int)
+
+    # TODO: should x, y be centroids? point on surface?
+    #       lat, long coordinates
+    x = np.array(query_result[0]['x'])
+    y = np.array(query_result[0]['y'])
+    coords = np.array(zip(x, y))
+
+    # extract dependent variable
+    Y = np.array(query_result[0]['dep_var']).reshape((-1, 1))
+
+    n = Y.shape[0]
+    k = len(ind_vars)
+    X = np.zeros((n, k))
+
+    for attr in range(0, k):
+        attr_name = 'attr' + str(attr + 1)
+        X[:, attr] = np.array(
+          query_result[0][attr_name]).flatten()
+
+    # add intercept variable name
+    ind_vars.insert(0, 'intercept')
+
+    # split data into "training" and "test" for predictions
+    # create index to split based on null y values
+    train = np.where(Y != np.array(None))[0]
+    test = np.where(Y == np.array(None))[0]
+    if len(test) < 1:
+        plpy.error('No rows flagged for prediction: verify that rows denoting'
+                'prediction locations have a dependent variable value of Null')
+
+    # split dependent variable (only need training which is non-Null's)
+    Y_train = Y[train].reshape((-1,1))
+    Y_train = Y_train.astype(np.float)
+
+    # split coords
+    coords_train = coords[train]
+    coords_test = coords[test]
+
+    # split explanatory variables
+    X_train = X[train]
+    X_test = X[test]
+
+    # calculate bandwidth if none is supplied
+    if bw is None:
+        bw = Sel_BW(coords_train, Y_train, X_train,
+                    fixed=fixed, kernel=kernel).search()
+    
+    # estimate model and predict at new locations
+    model = GWR(coords_train, Y_train, X_train, bw,
+                fixed=fixed, kernel=kernel).predict(coords_test, X_test)
+
+    coefficients = []
+    stand_errs = []
+    t_vals = []
+    r_squared = model.localR2.flatten()
+    predicted = model.predy.flatten()
+   
+    m = len(model.predy)
+    for idx in xrange(m):
+        coefficients.append(json.dumps({var: model.params[idx, k]
+                                        for k, var in enumerate(ind_vars)}))
+        stand_errs.append(json.dumps({var: model.bse[idx, k]
+                                      for k, var in enumerate(ind_vars)}))
+        t_vals.append(json.dumps({var: model.tvalues[idx, k]
+                                  for k, var in enumerate(ind_vars)}))
+
+    return zip(coefficients, stand_errs, t_vals,
+               r_squared, predicted, rowid)