ENH: Add auto_correlation as function itself not only as plot.

[pokecheater]

Feature Type

• Adding new functionality to pandas

• Changing existing functionality in pandas

• Removing existing functionality in pandas

Problem Description

Dear Pandas Team,

I hope this message finds you well. I am reaching out to request a new feature that I believe would be incredibly useful for many users. Currently, the autocorrelation_plot function is available in Pandas, but it is only accessible as a plot. For users who need to automate their data analysis workflows, it would be extremely valuable to have access to the autocorrelation values directly through a method.

I recently found myself in this exact situation and ended up copying the code from the plot function and deleting the plot section to obtain the values I needed. However, I believe that having a dedicated method to calculate the autocorrelation values would greatly streamline this process and save users a lot of time.

I understand that the Pandas Team is likely very busy, but I wanted to bring this idea to your attention in case it is something that could be implemented relatively easily. I have included the code I used below in the Feature Description to make it easier for you to understand my request.

Thank you for considering my suggestion. I greatly appreciate all of the hard work that goes into developing and maintaining such a powerful and versatile data analysis library.

Best regards

Feature Description

def auto_correlation(series: pd.DataFrame, column:str):
"""
Uses code from pandas to calculate the same auto_correlation as it is done in the plot auto_correlation method.

Args:
    series (pd.DataFrame): The pd.DataFrame to auto_correlate.
    column (str): The column to auto_correlate.

Returns:
    pd.DataFrame: The autocorrelated values inside a DataFrame. Corr column will contain the correlation value and lag the corresponding lag value. 
"""
series = series[column]
# SRC: https://github.com/pandas-dev/pandas/blob/2e218d10984e9919f0296931d92ea851c6a6faf5/pandas/plotting/_matplotlib/misc.py#L447
n = len(series)
data = np.asarray(series)
mean = np.mean(data)
c0 = np.sum((data - mean) ** 2) / n

def r(h):
    return ((data[: n - h] - mean) * (data[h:] - mean)).sum() / n / c0

x = np.arange(n) + 1
y = [r(loc) for loc in x]
return pd.DataFrame({"corr": y, "lag": x})

Alternative Solutions

null

Additional Context

No response

[DeaMariaLeon]

Hi @pokecheater, thank you for opening an issue with your suggestion.

I will (re)label it to "needs discussion", as core developers opinions are needed.

In the meantime, maybe this is useful to you:
https://pandas.pydata.org/docs/reference/api/pandas.Series.autocorr.html

This link is a bit old, but migh give you some ideas:
https://stackoverflow.com/questions/26083293/calculating-autocorrelation-of-pandas-dataframe-along-each-column

[mroeschke]

Thanks for the issue, but it appears this hasn't gotten traction in a while so closing