Pandas df.resample with column-specific aggregation function

Question

With pandas.DataFrame.resample I can downsample a DataFrame:

df.resample("3s", how="mean")

This resamples a data frame with a datetime-like index such that all values within 3 seconds are aggregated into one row. The values of the columns are averaged.

Question: I have a data frame with multiple columns. Is it possible to specify a different aggregation function for different columns, e.g. I want to "sum" column x, "mean" column y and pick the "last" for column z? How can I achieve that effect?

I know I could create a new empty data frame, and then call resample three times, but I would prefer a faster in-place solution.

Answer 1

You can use .agg after resample. With a dictionary, you can aggregate different columns with various functions.

Try this:

df.resample("3s").agg({'x':'sum','y':'mean','z':'last'})

Also, how is deprecated:

C:\Program Files\Anaconda3\lib\site-packages\ipykernel__main__.py:1: FutureWarning: how in .resample() is deprecated the new syntax is .resample(...).mean()

Answer 2

Consider the dataframe df

np.random.seed([3,1415])
tidx = pd.date_range('2017-01-01', periods=18, freq='S')
df = pd.DataFrame(np.random.rand(len(tidx), 3), tidx, list('XYZ'))
print(df)

                            X         Y         Z
2017-01-01 00:00:00  0.444939  0.407554  0.460148
2017-01-01 00:00:01  0.465239  0.462691  0.016545
2017-01-01 00:00:02  0.850445  0.817744  0.777962
2017-01-01 00:00:03  0.757983  0.934829  0.831104
2017-01-01 00:00:04  0.879891  0.926879  0.721535
2017-01-01 00:00:05  0.117642  0.145906  0.199844
2017-01-01 00:00:06  0.437564  0.100702  0.278735
2017-01-01 00:00:07  0.609862  0.085823  0.836997
2017-01-01 00:00:08  0.739635  0.866059  0.691271
2017-01-01 00:00:09  0.377185  0.225146  0.435280
2017-01-01 00:00:10  0.700900  0.700946  0.796487
2017-01-01 00:00:11  0.018688  0.700566  0.900749
2017-01-01 00:00:12  0.764869  0.253200  0.548054
2017-01-01 00:00:13  0.778883  0.651676  0.136097
2017-01-01 00:00:14  0.544838  0.035073  0.275079
2017-01-01 00:00:15  0.706685  0.713614  0.776050
2017-01-01 00:00:16  0.542329  0.836541  0.538186
2017-01-01 00:00:17  0.185523  0.652151  0.746060

Use agg

df.resample('3S').agg(dict(X='sum', Y='mean', Z='last'))

                            X         Y         Z
2017-01-01 00:00:00  1.760624  0.562663  0.777962
2017-01-01 00:00:03  1.755516  0.669204  0.199844
2017-01-01 00:00:06  1.787061  0.350861  0.691271
2017-01-01 00:00:09  1.096773  0.542220  0.900749
2017-01-01 00:00:12  2.088590  0.313316  0.275079
2017-01-01 00:00:15  1.434538  0.734102  0.746060