rolling skew and rolling kurt has wrong answer under all eq values · Issue #18044 · pandas-dev/pandas (original) (raw)

Code Sample, a copy-pastable example if possible

pd.Series([1.1]*15).rolling(10).skew() pd.Series([1.1]*15).rolling(10).kurt()

Problem description

rolling skew() wrong result:

pd.Series([1.1]*15).rolling(10).skew()

0 NaN
1 NaN
2 NaN
3 NaN
4 NaN
5 NaN
6 NaN
7 NaN
8 NaN
9 -1.056765e+08
10 -1.056765e+08
11 -1.056765e+08
12 -1.056765e+08
13 -1.056765e+08
14 -1.056765e+08
dtype: float64

rolling kurt() wrong result:
0 NaN
1 NaN
2 NaN
3 NaN
4 NaN
5 NaN
6 NaN
7 NaN
8 NaN
9 1.201335e+16
10 1.201335e+16
11 1.201335e+16
12 1.201335e+16
13 1.201335e+16
14 1.201335e+16
dtype: float64

But try pd.Series([1.1]*15).tail(10).skew() and pd.Series([1.1]*15).tail(10).kurt() could get the right result.
and also a reference workaround by using apply to call scipy:

pd.Series([1.1]*15).rolling(10).apply(lambda x: scipy.stats.skew(x)) pd.Series([1.1]*15).rolling(10).apply(lambda x: scipy.stats.kurt(x))

seems it has bug in rolling_skew/rolling_kurt

Expected Output

Output of pd.show_versions()

[paste the output of pd.show_versions() here below this line]

INSTALLED VERSIONS

commit: None
python: 3.5.2.final.0
python-bits: 64
OS: Linux
OS-release: 4.4.0-93-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8

pandas: 0.21.0
pytest: None
pip: 9.0.1
setuptools: 36.6.0
Cython: 0.24
numpy: 1.13.3
scipy: 1.0.0
pyarrow: None
xarray: None
IPython: 6.2.1
sphinx: None
patsy: 0.4.1
dateutil: 2.6.1
pytz: 2017.3
blosc: None
bottleneck: 1.2.0
tables: 3.2.2
numexpr: 2.5.2
feather: None
matplotlib: 2.0.0
openpyxl: 2.2.0-b1
xlrd: 0.9.4
xlwt: None
xlsxwriter: 0.7.3
lxml: 3.5.0
bs4: None
html5lib: 1.0b10
sqlalchemy: 1.0.12
pymysql: None
psycopg2: None
jinja2: 2.9.6
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None