conda install jupyterlab
# Under some unresolved circumstances pd.to_datetime()
# yields an object. This ensures the desired type.
pd.to_datetime(ts, utc=True).astype('datetime64[ns]')
# Compare two dataframes.
np.allclose(one, two)
Multi indexing:
alpha
date asset
2019-03-20 A 0.000000
B 0.920731
C 0.703412
D 0.596373
E 0.611381
... ...
2021-05-27 V -0.010958
W 1.676388
X 1.445200
Y 0.952216
Z 1.646791
[20800 rows x 1 columns]
# use .loc with multi-index
alpha.loc[(slice('2020-01-01', '2020-02-01'), slice(None)),:]
alpha
date asset
2020-01-01 A -0.713335
B -0.458934
-----------------
# TODO: What the hell this thing does?
same = 'a'
diff = 'b'
ys = xs[xs.duplicated(subset=[same], keep=False)]
ys[~ys.duplicated(subset=[diff], keep=False)]
C 0.072503
D 0.230143
E -0.354242
... ...
2020-02-01 V 2.790067
W 1.998978
X 3.426980
Y 2.649550
Z 2.438424
[832 rows x 1 columns]
# Winsorize
xs = xs.clip(lower=xs.quantile(0.01), upper=xs.quantile(0.99))
# Enfore types between dataframes.
b = b.astype(a.dtypes.to_dict())