Rolling averages on groups

Question

I've got a Series of the form:

Contract  Date      
196012    1960-01-05    110.70
          1960-01-07    110.70
          1960-01-08    110.40
          1960-01-11    110.00
          1960-01-12    109.60
          1960-01-13    109.70
          1960-01-14    109.50
          1960-01-15    109.60
          1960-01-18    109.60
          1960-01-19    110.20
          1960-01-20    110.00
          1960-01-21    110.30
          1960-01-22    110.00
          1960-01-25    109.50
          1960-01-26    109.60
          1960-01-28    109.70
          1960-01-29    110.00
          1960-02-01    109.60
          1960-02-02    109.60
          1960-02-03    109.60
          1960-02-04    110.10
          1960-02-05    110.20
          1960-02-08    110.20
          1960-02-09    110.50
          1960-02-10    110.10
          1960-02-11    109.50
          1960-02-12    110.40
          1960-02-15    110.00
          1960-02-16    109.50
          1960-02-17    110.00
                         ...  
201812    2016-06-29    403.00
          2016-06-30    398.00
          2016-07-01    404.25
          2016-07-05    402.00
          2016-07-06    394.00
201912    2015-12-16    417.00
          2015-12-17    415.00
          2015-12-23    416.00

Where the index is a hierarchical multi-index. I'd like to apply a rolling window average to each contract, and save the results to a new column.

What's the right way?

Perhaps see [this thread](http://stackoverflow.com/questions/13728392/moving-average-or-running-mean) — nbryans, Jul 12 '16 at 16:55
The issue isn't the rolling mean- it's the groups/multi-index that's the tricky part — cjm2671, Jul 12 '16 at 16:59

score 2 · Accepted Answer · answered Jul 12 '16 at 17:04

2

You want to groupby(level=0) then rolling(n).mean()

s.groupby(level=0).rolling(10).mean()

answered Jul 12 '16 at 17:04

piRSquared

285,575
57
475
624

The `Contract` level is duplicated. You can deduplicate it with `t.index = t.index.droplevel(0)`, where `t` is the result of your `groupby`. – Alicia Garcia-Raboso Jul 12 '16 at 17:14

score 1 · Answer 2 · answered Jul 12 '16 at 17:24

1

If you want the rolling mean to be a new column in the same dataframe:

df['rolling_mean'] = df.groupby(level=0).rolling(window_size).mean().values

answered Jul 12 '16 at 17:24

Grr

15,553
7
65
85

Rolling averages on groups

2 Answers2

Linked