BUG: mean overflows for integer dtypes (fixes pandas-dev#10155)

Author: mortada

pandas/core/nanops.py

@@ -254,7 +254,7 @@ def nansum(values, axis=None, skipna=True):
 @bottleneck_switch()
 def nanmean(values, axis=None, skipna=True):
     values, mask, dtype, dtype_max = _get_values(values, skipna, 0)
-    the_sum = _ensure_numeric(values.sum(axis, dtype=dtype_max))
+    the_sum = _ensure_numeric(values.sum(axis, dtype=np.float64))
     count = _get_counts(mask, axis)
 
     if axis is not None and getattr(the_sum, 'ndim', False):

pandas/tests/test_nanops.py

@@ -323,6 +323,17 @@ def test_nanmean(self):
                         allow_complex=False, allow_obj=False,
                         allow_str=False, allow_date=False, allow_tdelta=True)
 
+        # GH 10155
+        # In the previous implementation mean can overflow for int dtypes, it
+        # is now consistent with numpy
+        from pandas import Series
+        for a in [2 ** 55, -2 ** 55, 20150515061816532]:
+            s = Series(a, index=range(500))
+            result = s.mean()
+            np_result = s.values.mean()
+            self.assertEqual(result, a)
+            self.assertEqual(result, np_result)
+
     def test_nanmedian(self):
         self.check_funs(nanops.nanmedian, np.median,
                         allow_complex=False, allow_str=False, allow_date=False,