ENH:AttributeError: 'SeriesGroupBy' object has no attribute 'kurtosis'

[rsuhada]

Code Sample, a copy-pastable example

df = pd.DataFrame({'birb':['Falcon', 'Falcon', 'Falcon', 'Falcon',
                           'Parrot', 'Parrot', 'Parrot', 'Parrot'],
                   'value': [390., 350.,390., 350.,
                             30., 20., 30., 20.]})
df.groupby('birb').aggregate({'value': ['count', 'mean', 'skew']})

Problem description

Code gives error: AttributeError: 'SeriesGroupBy' object has no attribute 'kurtosis' instead of correct group-wise kurtosis.
Skew works ok.

Expected Output

       value            
       count   mean kurtosis
birb                    
Falcon     4  370.0  -6.0
Parrot     4   25.0  -6.0

Output of pd.show_versions()

INSTALLED VERSIONS

commit : 7d32926
python : 3.7.5.final.0
python-bits : 64
OS : Linux
OS-release : 5.4.0-66-generic
Version : #74~18.04.2-Ubuntu SMP Fri Feb 5 11:17:31 UTC 2021
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_IE.UTF-8
LOCALE : en_IE.UTF-8

pandas : 1.2.2
numpy : 1.17.5
pytz : 2021.1
dateutil : 2.8.1
pip : 21.0.1
setuptools : 49.2.1
Cython : None
pytest : 5.3.5
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 2.11.2
IPython : 7.20.0
pandas_datareader: None
bs4 : None
bottleneck : None
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : 3.3.4
numexpr : None
odfpy : None
openpyxl : 3.0.6
pandas_gbq : None
pyarrow : 3.0.0
pyxlsb : None
s3fs : None
scipy : 1.4.1
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : 0.52.0

[rhshadrach]

I don't believe SeriesGroupBy is documented as having a kurtosis method. Are you finding something different?

[rsuhada]

You are right... it just tripped me up that first three moments are implemented, but not the fourth one.
One can workaround it with the scipy implementation of kurtosis, but it would be nice to have it directly here.

[rhshadrach]

+1 on adding, I think this would just be adding to the common_apply_allowlist. In the meantime, from https://stackoverflow.com/questions/37345493/kurtosis-on-groupby-of-pandas-dataframe-doesnt-work

df.groupby('a').apply(pd.DataFrame.kurt)

This is essentially what is occurring under the hood for skew. Might want to use pd.Series instead of pd.DataFrame here.

[jaepil-choi]

Can't believe this is still a thing. mean, std, skew, count, min, max, median all other things work but not kurtosis? I mean.. why?

[rhshadrach]

@jaepil-choi - PRs are welcome!

[jaepil-choi]

@rhshadrach

Found out that pd.DataFrame.kurt workaround works. Would a very basic temporary fix like simply converting 'kurt' to pd.DataFrame.kurt get passed and merged?

[rhshadrach]

I believe you can implement this similar to corr:

SeriesGroupBy:

pandas/pandas/core/groupby/generic.py

Lines 1413 to 1423 in c47296a

	@doc(Series.corr.__doc__)
	def corr(
	self,
	other: Series,
	method: CorrelationMethod = "pearson",
	min_periods: int | None = None,
	) -> Series:
	result = self._op_via_apply(
	"corr", other=other, method=method, min_periods=min_periods
	)
	return result

DataFrameGroupBy:

pandas/pandas/core/groupby/generic.py

Lines 2894 to 2904 in c47296a

	@doc(DataFrame.corr.__doc__)
	def corr(
	self,
	method: str | Callable[[np.ndarray, np.ndarray], float] = "pearson",
	min_periods: int = 1,
	numeric_only: bool = False,
	) -> DataFrame:
	result = self._op_via_apply(
	"corr", method=method, min_periods=min_periods, numeric_only=numeric_only
	)
	return result

A PR would need the right arguments and tests.

[saldanhad]

take

[Shubhank-Gyawali]

take

[snitish]

take

[snitish]

@rhshadrach I raised a PR for this - #60433. Could you please review? Thanks.