如何从Pandas的时间序列的最后四个完整季度中选择数据？

Question

Say I've got a dataframe with a datetime index which covers the last financial year and one day in the current financial year (starting on April 1) :

           Units
date
2016-01-01   8734   
2016-06-30   6120
2016-09-30   7346
2016-12-31   5925
2016-03-31   7542
2016-06-30   9916
2016-09-30   9547
2016-12-31   8063
2017-01-01   7000
2017-03-31   5672
2017-04-01   7856

I'd like to be able to select the data for the last complete four quarters - in this case ignoring the first and last rows.

I know I can do this with slicing, thus:

df["2016-04-01":"2017-03-31"]

What's the most elegant - pythonic - solution to filter the data according to the last four complete quarters programmatically?

Answer 1

You should first define your quarters. You can use pd.period_range for that with the correct freq : example :

quarters = pd.period_range('2016Q1', '2017Q1', freq='Q-MAR')

This would give you a PeriodIndex on which you can change the frequency to get the dates you want with asfreq :

quarters.asfreq('D', 'E')

That would give you the PeriodIndex that you can use to slice your Index.

Here are more example in the documentation .

Answer 2

pandas.DatetimeIndex.quarter Might also be useful.

And then you can use groupby to aggregate easily.

Answer 3

Using Alex's pointer to the DateOffset functionality in Pandas I found a partial solution, as well as the datetime module:

import datetime
from pandas.tseries.offsets import *
now = datetime.datetime.now()
start_year = (now - BQuarterEnd(n=1) - (12 * MonthBegin())).to_datetime()
end_year = (now - BQuarterEnd(n=1) ).to_datetime()
df[start_year.strftime("%Y-%m-%d") : end_year.strftime("%Y-%m-%d")]