CODEWITHSUNDEEP
pythonpandasstatisticssignal-processingstatsmodels
s5s
s5s

Reputation: 12134

Calculate autocorrelation as a function of lag in Python

In python (+ pandas/numpy/scipy/statsmodels) Is there a function to return the autocorrelations wrt lag? Does anything like this exist ready made as a library function?

To avoid confusion, I want the following, just I don't want to plot it but I want it returned as a series (pd.Series or pd.DataFrame):

import numpy as np
import pandas as pd
from statsmodels.graphics.tsaplots import plot_acf
from matplotlib import pyplot as plt
plt.ion()
s = pd.Series(np.sin(range(1,100))) + pd.Series(np.random.randn(99))
plot_acf(s)

Effectively, I want what pd.Series.autocorr() returns but I want a series rather than a scalar returned where the series contains the autocorrelation for various lags.

Edit:

One way to achieve the above is to do:

pd.Series([s.autocorr(i) for i in range(0,s.shape[0]-1)], index=range(0,s.shape[0]-1))

Upvotes: 5

Views: 8728

Answers (2)

cfulton
cfulton

Reputation: 3195

How about the Statsmodels acf function?

import statsmodels.api as sm

np.random.seed(1234)
s = pd.Series(np.sin(range(1,100))) + pd.Series(np.random.randn(99))
pd.Series(sm.tsa.acf(s, nlags=5))

yields

0    1.000000
1    0.033136
2   -0.124275
3   -0.396403
4   -0.248519
5    0.078170
dtype: float64

Upvotes: 5

BENY
BENY

Reputation: 323226

I can only think of speed up your own method vectorize

np.vectorize(s.autocorr)(np.arange(0,len(s)-1))

Upvotes: 0

Related Questions