Using resample to align multiple timeseries in pandas

问题

Here's the setup code:

import pandas
from datetime import datetime

a_values = [1728, 1635, 1733]
a_index = [datetime(2011, 10, 31), datetime(2012, 1, 31), datetime(2012, 4, 30)]
a = pandas.Series(data=a_values, index=a_index)

aa_values = [6419, 5989, 6006]
aa_index = [datetime(2011, 9, 30), datetime(2011, 12, 31), datetime(2012, 3, 31)]
aa = pandas.Series(data=aa_values, index=aa_index)

apol_values = [1100, 1179, 969]
apol_index = [datetime(2011, 8, 31), datetime(2011, 11, 30), datetime(2012, 2, 29)]
apol = pandas.Series(data=apol_values, index=apol_index)

Here's what the data looks like in a table (3rd value for APOL isn't shown):

The goal is to align the data to calendar quarter markers so the 3 data sets can be compared. Just glancing at the below dates, Mar 2012, Dec 2011, and Sep 2011 seem like reasonable markers for alignment.

Here's the output with fill_method='ffill':

In [6]: a.resample('Q', fill_method='ffill')
Out[6]: 
2011-12-31    1728
2012-03-31    1635
2012-06-30    1733
Freq: Q-DEC

In [7]: aa.resample('Q', fill_method='ffill')
Out[7]: 
2011-09-30    6419
2011-12-31    5989
2012-03-31    6006
Freq: Q-DEC

In [8]: apol.resample('Q', fill_method='ffill')
Out[8]: 
2011-09-30    1100
2011-12-31    1179
2012-03-31     969
Freq: Q-DEC

Which looks like this:

Notice how the most recent numbers in each series don't line up.

And here's the output with fill_method='bfill':

In [9]: a.resample('Q', fill_method='bfill')
Out[9]: 
2011-12-31    1635
2012-03-31    1733
2012-06-30     NaN
Freq: Q-DEC

In [10]: aa.resample('Q', fill_method='bfill')
Out[10]: 
2011-09-30    6419
2011-12-31    5989
2012-03-31    6006
Freq: Q-DEC

In [11]: apol.resample('Q', fill_method='bfill')
Out[11]: 
2011-09-30    1179
2011-12-31     969
2012-03-31     NaN
Freq: Q-DEC

Which looks like this:

Again, the most recent numbers in the series don't line up.

Is this the expected output of resample() in this scenario?

What can I do to get results where the most recent 3 numbers above are aligned and everything else follows appropriately?

EDIT: Here's what the desired output looks like:

回答1:

df1 = DataFrame({'a':a})
df2 = DataFrame({'aa':aa})
df3 = DataFrame({'apol':apol})
df=df1.append([df2,df3]).sort_index()
print df.resample('Q-APR',loffset='-1m').T

Output:

      2011-09-30  2011-12-31  2012-03-31
a           1728        1635        1733
aa          6419        5989        6006
apol        1100        1179         969

来源：https://stackoverflow.com/questions/13185454/using-resample-to-align-multiple-timeseries-in-pandas

标签

python

time-series

pandas

resampling