BUG: exponential moving window covariance fails for multiIndexed DataFrame

[PablocFonseca]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Note: Please read this guide detailing how to provide the necessary information for us to reproduce your bug.

Code Sample, a copy-pastable example

import pandas as pd
import numpy as np 

columns = pd.MultiIndex.from_product([['a','b','c'],['x','y','w','z'], list(range(9))])
index = range(1000)
df = pd.DataFrame(
    np.random.normal(size=(len(index), len(columns))),
    index=index,
    columns=columns
    )
    
df.ewm(alpha=0.1).cov()  #Throws AssertionError: Length of order must be same as number of levels (4), got 3

Problem description

When calculating ewm covariance, pandas fails when the DataFrame has multiindex columns. However it works when columns are simple Index dataframes.
It works for:

pd.DataFrame(df.values).ewm(alpha=0.1).cov()

Expected Output

The covariance, actually only the last matrix (last level of index)

Output of pd.show_versions()

INSTALLED VERSIONS

commit : None
python : 3.7.7.final.0
python-bits : 64
OS : Windows
OS-release : 10
machine : AMD64
processor : Intel64 Family 6 Model 142 Stepping 10, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : None.None

pandas : 1.0.3
numpy : 1.18.1
pytz : 2019.3
dateutil : 2.8.1
pip : 20.0.2
setuptools : 46.1.3.post20200330
Cython : 0.29.15
pytest : 5.4.1
hypothesis : 5.8.3
sphinx : 2.4.4
blosc : None
feather : None
xlsxwriter : 1.2.8
lxml.etree : 4.5.0
html5lib : 1.0.1
pymysql : None
psycopg2 : None
jinja2 : 2.11.1
IPython : 7.13.0
pandas_datareader: None
bs4 : 4.9.0
bottleneck : 1.3.2
fastparquet : None
gcsfs : None
lxml.etree : 4.5.0
matplotlib : 3.1.3
numexpr : 2.7.1
odfpy : None
openpyxl : 3.0.3
pandas_gbq : None
pyarrow : 0.15.1
pytables : None
pytest : 5.4.1
pyxlsb : None
s3fs : None
scipy : 1.4.1
sqlalchemy : 1.3.16
tables : 3.6.1
tabulate : 0.8.3
xarray : None
xlrd : 1.2.0
xlwt : 1.3.0
xlsxwriter : 1.2.8
numba : 0.49.0

[arw2019]

Confirming that this bug exists in the master version of pandas.

Output of pd.show_versions()

INSTALLED VERSIONS

commit : 62c7dd3
python : 3.8.2.final.0
python-bits : 64
OS : Linux
OS-release : 4.15.0-101-generic
Version : #102-Ubuntu SMP Mon May 11 10:07:26 UTC 2020
machine : x86_64
processor :
byteorder : little
LC_ALL : C.UTF-8
LANG : C.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.1.0.dev0+1681.g62c7dd3e7.dirty
numpy : 1.17.5
pytz : 2020.1
dateutil : 2.8.1
pip : 20.1.1
setuptools : 46.4.0.post20200518
Cython : 0.29.19
pytest : 5.4.2
hypothesis : 5.15.1
sphinx : 3.0.4
blosc : None
feather : None
xlsxwriter : 1.2.8
lxml.etree : 4.5.1
html5lib : 1.0.1
pymysql : None
psycopg2 : None
jinja2 : 2.11.2
IPython : 7.14.0
pandas_datareader: None
bs4 : 4.9.1
bottleneck : 1.3.2
fastparquet : 0.4.0
gcsfs : None
matplotlib : 3.2.1
numexpr : 2.7.1
odfpy : None
openpyxl : 3.0.3
pandas_gbq : None
pyarrow : 0.17.1
pytables : None
pyxlsb : None
s3fs : 0.4.2
scipy : 1.4.1
sqlalchemy : 1.3.17
tables : 3.6.1
tabulate : 0.8.7
xarray : 0.15.1
xlrd : 1.2.0
xlwt : 1.3.0
numba : 0.49.1

[arw2019]

For the purposes of debugging I tweaked @PablocFonseca's snippet to make the DataFrames are smaller and that way easier to inspect. I believe the snippet below throws the same error as the original report.

import pandas as pd
import numpy as np 

columns = pd.MultiIndex.from_product([['a', 'b'],['x','y'], list(range(2))])
index = range(3)
df = pd.DataFrame(
    np.random.normal(size=(len(index), len(columns))),
    index=index,
    columns=columns
    )
    
df.ewm(alpha=0.1).cov()  #Throws AssertionError: Length of order must be same as number of levels (4), got 3

[arw2019]

I tracked down the origin of the bug to the _flex_binary_moment method in pandas/core/window/common.py.

I haven't figured out the full details but I think that by the time of line 182:

result = result.reorder_levels([2, 0, 1]).sort_index()   # line 182

the DataFrame result has been incorrectly indexed. For my example the levels of result's MultiIndex are:

levels=[Index(['a', 'b'], dtype='object'), Index(['x', 'y'], dtype='object'), Int64Index([0, 1], dtype='int64'), Int64Index([0, 1, 2], dtype='int64')]

and then obviously reorder_levels fails in line 182 because that assumes that result has only three levels in its MultiIndex.

I haven't understood _flex_binary_moment well enough to quite know what the correct behavior is and what adjustment is needed - but I'll keep digging!

[jorisvandenbossche]

@PablocFonseca thanks for the report, and @arw2019 thanks for the confirmation and simple reproducer!

It's also failing on 0.25

[arw2019]

take

[arw2019]

Would love your feedback on #34943 - I think I fixed this problem there

[Dinkarkumar]

I want to work on this issue.