diff --git a/_unittests/ut_df/test_pandas_groupbynan.py b/_unittests/ut_df/test_pandas_groupbynan.py
index 141568b..63af8fc 100644
--- a/_unittests/ut_df/test_pandas_groupbynan.py
+++ b/_unittests/ut_df/test_pandas_groupbynan.py
@@ -1,3 +1,4 @@
+# coding: utf-8
 """
 @brief      test log(time=1s)
 """
@@ -5,7 +6,7 @@
 import pandas
 import numpy
 from scipy.sparse.linalg import lsqr as sparse_lsqr
-from pyquickhelper.pycode import ExtTestCase
+from pyquickhelper.pycode import ExtTestCase, ignore_warnings
 from pandas_streaming.df import pandas_groupby_nan, numpy_types
 
 
@@ -102,6 +103,40 @@ def test_pandas_groupbynan_regular_nanback(self):
             lambda: pandas_groupby_nan(df, ["a", "cc"], nanback=True).sum(),
             NotImplementedError)
 
+    def test_pandas_groupbynan_doc(self):
+        data = [dict(a=2, ind="a", n=1),
+                dict(a=2, ind="a"),
+                dict(a=3, ind="b"),
+                dict(a=30)]
+        df = pandas.DataFrame(data)
+        gr2 = pandas_groupby_nan(df, ["ind"]).sum()
+        ind = list(gr2['ind'])
+        self.assertTrue(numpy.isnan(ind[-1]))
+        val = list(gr2['a'])
+        self.assertEqual(val[-1], 30)
+
+    @ignore_warnings(UserWarning)
+    def test_pandas_groupbynan_doc2(self):
+        data = [dict(a=2, ind="a", n=1),
+                dict(a=2, ind="a"),
+                dict(a=3, ind="b"),
+                dict(a=30)]
+        df = pandas.DataFrame(data)
+        gr2 = pandas_groupby_nan(df, ["ind", "a"], nanback=False).sum()
+        ind = list(gr2['ind'])
+        self.assertEqual(ind[-1], "²nan")
+
+    def test_pandas_groupbynan_doc3(self):
+        data = [dict(a=2, ind="a", n=1),
+                dict(a=2, ind="a"),
+                dict(a=3, ind="b"),
+                dict(a=30)]
+        df = pandas.DataFrame(data)
+        self.assertRaise(lambda: pandas_groupby_nan(df, ["ind", "n"]).sum(),
+                         NotImplementedError)
+        # ind = list(gr2['ind'])
+        # self.assertTrue(numpy.isnan(ind[-1]))
+
 
 if __name__ == "__main__":
     unittest.main()
diff --git a/pandas_streaming/df/dataframe_helpers.py b/pandas_streaming/df/dataframe_helpers.py
index 18ead2e..7591fdf 100644
--- a/pandas_streaming/df/dataframe_helpers.py
+++ b/pandas_streaming/df/dataframe_helpers.py
@@ -289,7 +289,7 @@ def pandas_fillna(df, by, hasna=None, suffix=None):
     :param suffix: use a prefix for the NaN value
     :return: list of values chosen for each column, new dataframe (new copy)
     """
-    suffix = suffix if suffix else "²"
+    suffix = suffix if suffix else "²nan"
     df = df.copy()
     rep = {}
     for c in by:
@@ -364,7 +364,10 @@ def pandas_groupby_nan(df, by, axis=0, as_index=False, suffix=None, nanback=True
 
             from pandas import DataFrame
 
-            data = [dict(a=2, ind="a", n=1), dict(a=2, ind="a"), dict(a=3, ind="b"), dict(a=30)]
+            data = [dict(a=2, ind="a", n=1),
+                    dict(a=2, ind="a"),
+                    dict(a=3, ind="b"),
+                    dict(a=30)]
             df = DataFrame(data)
             print(df)
             gr = df.groupby(["ind"]).sum()
@@ -378,7 +381,10 @@ def pandas_groupby_nan(df, by, axis=0, as_index=False, suffix=None, nanback=True
             from pandas import DataFrame
             from pandas_streaming.df import pandas_groupby_nan
 
-            data = [dict(a=2, ind="a", n=1), dict(a=2, ind="a"), dict(a=3, ind="b"), dict(a=30)]
+            data = [dict(a=2, ind="a", n=1),
+                    dict(a=2, ind="a"),
+                    dict(a=3, ind="b"),
+                    dict(a=30)]
             df = DataFrame(data)
             gr2 = pandas_groupby_nan(df, ["ind"]).sum()
             print(gr2)
@@ -436,10 +442,22 @@ def pandas_groupby_nan(df, by, axis=0, as_index=False, suffix=None, nanback=True
                         res.grouper.groupings[0].grouping_vector = arr
                         if (hasattr(res.grouper.groupings[0], '_cache') and
                                 'result_index' in res.grouper.groupings[0]._cache):
-                            res.grouper.groupings[0]._cache = {}
+                            index = res.grouper.groupings[0]._cache['result_index']
+                            if len(rep) == 1:
+                                key = list(rep.values())[0]
+                                new_index = numpy.array(index)
+                                for i in range(0, len(new_index)):  # pylint: disable=C0200
+                                    if new_index[i] == key:
+                                        new_index[i] = numpy.nan
+                                res.grouper.groupings[0]._cache['result_index'] = (
+                                    index.__class__(new_index))
+                            else:
+                                raise NotImplementedError(
+                                    "NaN values not implemented for multiindex.")
                     else:
-                        raise NotImplementedError("Not implemented for type: {0}".format(
-                            type(res.grouper.groupings[0].grouper)))
+                        raise NotImplementedError(
+                            "Not implemented for type: {0}".format(
+                                type(res.grouper.groupings[0].grouper)))
                 res.grouper._cache['result_index'] = res.grouper.groupings[0]._group_index
         else:
             if not nanback: