In pandas, how can I write a rolling window that has preceding as well as following rows?
for example, a sql rolling window can be written as follows:
sum(id_0) over (partition by id_1 order by id_2 rows between 3 preceding and 1 following)
i am not sure how to write this in pandas.
You can calculate the rolling sum for 4 preceding values then use the shift() method to apply those values to one row prior - for example:
rolling_ds = dataset.rolling(4, min_periods=1).sum().shift(-1)
A more robust discussion of this approach can be found here: Calculating Rolling forward averages with pandas
