[Python] most frequent *rolling* value on a object/str pandas dataframe column

I would like to have a new column with the most frequent rolling value from another str/object column.

date name state
0 2024-02-29 Alice CA
1 2024-02-27 Bob HI
2 2024-02-29 Cindy ID
3 2024-02-25 Dan MT
4 2024-02-29 Elliott CA
.. ... ... ...


I am trying to get the most frequent rolling state (for each row).

I have tried different combinations (and subsets) of

• .rolling()
• .apply()
• .mode()
• mode() from different libraries
• custom mode() function

which usually generates one of a handful of errors complaining that the column is non-numeric. I understand what the error is telling me - that it expects to aggregate and apply a numeric function (.mean() .sum() ...) - but its not even getting to the .apply() function...

def fail_now(x):
raise Exception('wow! we made it here!')


>>> df['state'].rolling(window=25).apply(fail_now)
...
pandas.errors.DataError: No numeric types to aggregate


>>> df[['state']].rolling(window=25).apply(fail_now)
...
pandas.errors.DataError: Cannot aggregate non-numeric type: object


>>> df[['state']].rolling(window=25)['state'].apply(fail_now)
...
pandas.errors.DataError: No numeric types to aggregate


I also tried a multitude of different things, including the raw flag in .apply() with no luck

Continue reading...