Question

I try to upsample daily TimeSeries values using the pandas resample function. When i’m upsampling a single day (2012-01-01) i expect the result to be the mean of the day considered for upsampling. The result should look like this:

2012-01-01   -0,285344
Freq: D

However, pandas returns two days like this:

2012-01-01   -0.412417
2012-01-02    0.127073
Freq: D

Is this a bug or a feature? If it is a feature how can i set the resample arguments to achieve my goal?

Thanks,
Thomas

Here is the full example:

In [66]: rng = date_range('1/1/2012', periods=24, freq='H')

In [67]: ts = Series(np.random.randn(len(rng)), index=rng)

In [68]: ts
Out[68]:
2012-01-01 00:00:00   -0.412417
2012-01-01 01:00:00    0.442482
2012-01-01 02:00:00    1.321009
2012-01-01 03:00:00    0.104408
2012-01-01 04:00:00    1.124611
2012-01-01 05:00:00   -1.041293
2012-01-01 06:00:00    1.194104
2012-01-01 07:00:00   -0.249706
2012-01-01 08:00:00    1.927320
2012-01-01 09:00:00   -0.828365
2012-01-01 10:00:00    0.163760
2012-01-01 11:00:00   -0.736053
2012-01-01 12:00:00   -0.323408
2012-01-01 13:00:00    1.478162
2012-01-01 14:00:00    1.449437
2012-01-01 15:00:00   -1.114443
2012-01-01 16:00:00   -0.003780
2012-01-01 17:00:00    0.554562
2012-01-01 18:00:00   -2.019614
2012-01-01 19:00:00    0.463484
2012-01-01 20:00:00    0.862818
2012-01-01 21:00:00   -1.280048
2012-01-01 22:00:00   -0.137987
2012-01-01 23:00:00   -0.428777
Freq: H

In [69]: ts.resample('D')
Out[69]:
2012-01-01   -0.412417
2012-01-02    0.127073
Freq: D

Answer 1

You could get the result you want by using the loffset and closed options.

>>> ts.resample('D', how='mean', closed="left", loffset=datetime.timedelta(days=-1))
2012-01-01    0.200299