I have monthly data. I want to convert it to "periods" of 3 months where q1 starts in January. So in the example below, the first 3 month aggregation would translate into start of q2 (desired format: 1996q2). And the data value that results from mushing together 3 monthly values is a mean (average) of 3 columns. Conceptually, not complicated. Does anyone know how to do it in one swoop? Potentially, I could do a lot of hard work through looping and just hardcode the hell out of it, but I am new to pandas and looking for something more clever than brute force.
1996-04 1996-05 1996-06 1996-07 ..... 25 19 37 40
So I am looking for:
1996q2 1996q3 1996q4 1997q1 1997q2 ..... avg avg avg ... ...
you can use pd.PeriodIndex(..., freq='Q') in conjunction with groupby(..., axis=1):
In [63]: df
Out[63]:
1996-04 1996-05 2000-07 2000-08 2010-10 2010-11 2010-12
0 1 2 3 4 1 1 1
1 25 19 37 40 1 2 3
2 10 20 30 40 4 4 5
In [64]: df.groupby(pd.PeriodIndex(df.columns, freq='Q'), axis=1).mean()
Out[64]:
1996Q2 2000Q3 2010Q4
0 1.5 3.5 1.000000
1 22.0 38.5 2.000000
2 15.0 35.0 4.333333
UPDATE: to get columns in a resulting DF as strings intead of period dtype:
In [66]: res = (df.groupby(pd.PeriodIndex(df.columns, freq='Q'), axis=1)
.mean()
.rename(columns=lambda c: str(c).lower()))
In [67]: res
Out[67]:
1996q2 2000q3 2010q4
0 1.5 3.5 1.000000
1 22.0 38.5 2.000000
2 15.0 35.0 4.333333
In [68]: res.columns.dtype
Out[68]: dtype('O')