BUG: Fix min_count issue for groupby.sum (pandas-dev#32914)

* Add test

* Check for null

* Release note

* Update and comment

* Update test

* Hack

* Try a different casting

* No pd

* Only for add

* Undo

* Release note

* Fix

* Space

* maybe_cast_result

* float -> Int

* Less if

Co-authored-by: Simon Hawkins <[email protected]>

Author: 2 people

doc/source/whatsnew/v1.1.0.rst

@@ -752,6 +752,7 @@ Groupby/resample/rolling
 - Bug in :meth:`DataFrameGroupby.transform` produces incorrect result with transformation functions (:issue:`30918`)
 - Bug in :meth:`GroupBy.count` causes segmentation fault when grouped-by column contains NaNs (:issue:`32841`)
 - Bug in :meth:`DataFrame.groupby` and :meth:`Series.groupby` produces inconsistent type when aggregating Boolean series (:issue:`32894`)
+- Bug in :meth:`DataFrameGroupBy.sum` and :meth:`SeriesGroupBy.sum` where a large negative number would be returned when the number of non-null values was below ``min_count`` for nullable integer dtypes (:issue:`32861`)
 - Bug in :meth:`SeriesGroupBy.quantile` raising on nullable integers (:issue:`33136`)
 - Bug in :meth:`SeriesGroupBy.first`, :meth:`SeriesGroupBy.last`, :meth:`SeriesGroupBy.min`, and :meth:`SeriesGroupBy.max` returning floats when applied to nullable Booleans (:issue:`33071`)
 - Bug in :meth:`DataFrameGroupBy.agg` with dictionary input losing ``ExtensionArray`` dtypes (:issue:`32194`)

pandas/core/groupby/ops.py

@@ -18,6 +18,7 @@
 from pandas.errors import AbstractMethodError
 from pandas.util._decorators import cache_readonly
 
+from pandas.core.dtypes.cast import maybe_cast_result
 from pandas.core.dtypes.common import (
     ensure_float64,
     ensure_int64,
@@ -548,17 +549,6 @@ def _cython_operation(
             if mask.any():
                 result = result.astype("float64")
                 result[mask] = np.nan
-        elif (
-            how == "add"
-            and is_integer_dtype(orig_values.dtype)
-            and is_extension_array_dtype(orig_values.dtype)
-        ):
-            # We need this to ensure that Series[Int64Dtype].resample().sum()
-            # remains int64 dtype.
-            # Two options for avoiding this special case
-            # 1. mask-aware ops and avoid casting to float with NaN above
-            # 2. specify the result dtype when calling this method
-            result = result.astype("int64")
 
         if kind == "aggregate" and self._filter_empty_groups and not counts.all():
             assert result.ndim != 2
@@ -582,6 +572,9 @@ def _cython_operation(
         elif is_datetimelike and kind == "aggregate":
             result = result.astype(orig_values.dtype)
 
+        if is_extension_array_dtype(orig_values.dtype):
+            result = maybe_cast_result(result=result, obj=orig_values, how=how)
+
         return result, names
 
     def aggregate(

pandas/tests/groupby/test_function.py

@@ -1014,3 +1014,20 @@ def test_apply_to_nullable_integer_returns_float(values, function):
     result = groups.agg([function])
     expected.columns = MultiIndex.from_tuples([("b", function)])
     tm.assert_frame_equal(result, expected)
+
+
+def test_groupby_sum_below_mincount_nullable_integer():
+    # https://github.com/pandas-dev/pandas/issues/32861
+    df = pd.DataFrame({"a": [0, 1, 2], "b": [0, 1, 2], "c": [0, 1, 2]}, dtype="Int64")
+    grouped = df.groupby("a")
+    idx = pd.Index([0, 1, 2], dtype=object, name="a")
+
+    result = grouped["b"].sum(min_count=2)
+    expected = pd.Series([pd.NA] * 3, dtype="Int64", index=idx, name="b")
+    tm.assert_series_equal(result, expected)
+
+    result = grouped.sum(min_count=2)
+    expected = pd.DataFrame(
+        {"b": [pd.NA] * 3, "c": [pd.NA] * 3}, dtype="Int64", index=idx
+    )
+    tm.assert_frame_equal(result, expected)