PERF: Series.corr/Series.cov for EA dtypes

[lukemanley]

• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/v2.1.0.rst file if fixing a bug or adding a new feature.

import pandas as pd
import numpy as np

N = 1_000_000

ser1 = pd.Series(np.random.randn(N), dtype="Float64")
ser2 = pd.Series(np.random.randn(N), dtype="Float64")

%timeit ser1.corr(ser2)

# 849 ms ± 56.1 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)    -> main
# 33.7 ms ± 320 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)  -> PR

%timeit ser1.cov(ser2)

# 739 ms ± 15.4 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)    -> main
# 32.9 ms ± 713 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)  -> PR

Also, aligns the error raised for non-numeric data with that of DataFrame.corr and DataFrame.cov.

[jbrockmendel]

any chance this gets weird if we have non-numeric?

is this the right thing to do if we have NAs?

[lukemanley]

any chance this gets weird if we have non-numeric?

This raises for non-numeric that cannot be cast to float (same as behavior as main branch):

ser = pd.Series(["a", "b"])
ser.corr(ser)

# ValueError: could not convert string to float: 'a'

It allows non-numeric that can be cast to float (same behavior as main branch):

ser = pd.Series(["1", "2"])
ser.corr(ser)

# 0.9999999999999999

is this the right thing to do if we have NAs?

NAs get dropped either way in the underlying nanops.py functions nancorr and nancov.

FYI, this is the same approach used in DataFrame.corr:

pandas/pandas/core/frame.py

Line 10381 in 0e39016

mat = data.to_numpy(dtype=float, na_value=np.nan, copy=False)

[jbrockmendel]

FYI, this is the same approach used in DataFrame.corr:

good point. works for me

[mroeschke]

Thanks @lukemanley