Python - 每组观察的 ACF 和 PACF plot

Question

I have a dataset (Python dataframe) that looks like:

Date      Group  Value
Date1     Group1 Value1
Date2     Group1 Value2
Date3     Group1 Value3
Date1     Group2 Value1
Date2     Group2 Value2
Date3     Group3 Value3

I would like to run a ACF and PACF plot on this dataset, but I don't want it to use the lags of different groups. Ie the 1st lag of Date1 Group2 Value1 should not be the previous value which belongs to group1, ie I want it to use the correct lags within groups.

How can I do that?

Answer 1

It's a little convoluted, but it would be possible to get separate ACF/ PACF like this:

dates = pd.date_range(df.index.min(), df.index.max(), freq="d") # assumes df.index is datetime

results_pacf = []
for group_name, df_group in df.groupby("group"):
        df_group = (pd.DataFrame(index=dates).merge(df_group, left_index=True, right_index=True, how="left").fillna(0))
        results_pacf.append({group_name: pacf(df_group.drop("group", axis=1))})`

results_pacf would be a list of dictionaries looking like [{group_name: [1, 0.22, -0.4, ...]}, {...}]