Merge pull request #145 from bashtage/futher-perf

ENH: Improve performance of PanelOLS

Author: bashtage

README.rst

@@ -105,10 +105,10 @@ The master branch can be installed by cloning the repo and running setup
 Documentation
 -------------
 
-`Stable Documentation <https://bashtage.github.io/linearmodels/>`__
-is built on every tagged version using
+`Stable Documentation <https://bashtage.github.io/linearmodels/>`__ is
+built on every tagged version using
 `doctr <https://github.com/drdoctr/doctr>`__. `Development
-Documentation <https://bashtage.github.io/linearmodels/devel/>`__ is
+Documentation <https://bashtage.github.io/linearmodels/devel>`__ is
 automatically built on every successful build of master.
 
 Plan and status
@@ -137,9 +137,9 @@ others are the version that are being used in the test environment. It
 is possible that older versions work.
 
 -  **Python 3.5+**: extensive use of ``@`` operator
--  NumPy (1.11+)
+-  NumPy (1.12+)
 -  SciPy (0.18+)
--  pandas (0.19+)
+-  pandas (0.20+)
 -  statsmodels (0.8+)
 -  xarray (0.9+, optional)
 

linearmodels/panel/data.py

@@ -35,9 +35,9 @@ def __init__(self, df):
         index = df.index
         self._major_axis = pd.Series(index.levels[1][index.labels[1]]).unique()
         self._minor_axis = pd.Series(index.levels[0][index.labels[0]]).unique()
-        full_index = list(product(self._minor_axis, self._major_axis))
-        self._full_index = pd.MultiIndex.from_tuples(full_index)
-        new_df = df.copy().loc[self._full_index]
+        self._full_index = pd.MultiIndex.from_product([self._minor_axis,
+                                                       self._major_axis])
+        new_df = df.reindex(self._full_index)
         self._frame = new_df
         i, j, k = len(self._items), len(self._major_axis), len(self.minor_axis)
         self._shape = (i, j, k)
@@ -507,7 +507,7 @@ def count(self, group='entity'):
         index = self.panel.minor_axis if group == 'entity' else self.panel.major_axis
         out = DataFrame(count.T, index=index, columns=self.vars)
         reindex = self.entities if group == 'entity' else self.time
-        out = out.loc[reindex].astype(np.int64)
+        out = out.reindex(reindex).astype(np.int64)
         out.index.name = group
         return out
 
@@ -550,7 +550,7 @@ def mean(self, group='entity', weights=None):
             mu = weighted_sum / sum_weights
 
         reindex = self.entities if group == 'entity' else self.time
-        out = mu.loc[reindex]
+        out = mu.reindex(reindex)
 
         return out
 

linearmodels/panel/model.py

@@ -370,8 +370,9 @@ def deferred_f():
 
     def _prepare_between(self):
         """Prepare values for between estimation of R2"""
-        y = self.dependent.mean('entity', weights=self.weights).values
-        x = self.exog.mean('entity', weights=self.weights).values
+        weights = self.weights if self._is_weighted else None
+        y = self.dependent.mean('entity', weights=weights).values
+        x = self.exog.mean('entity', weights=weights).values
         # Weight transformation
         wcount, wmean = self.weights.count('entity'), self.weights.mean('entity')
         wsum = wcount * wmean

linearmodels/utility.py

@@ -11,6 +11,7 @@
 
 from linearmodels.compat.numpy import lstsq
 
+
 class MissingValueWarning(Warning):
     pass