pandas allows for cool slicing on time indexes. For example, I can slice a dataframe df for the months from Janurary 2012 to March 2012 by doing:
df['2012-01':'2012-03']
However, I have a dataframe df with a multiindex where the time index is the second level. It looks like:
A B C D E
a 2001-01-31 0.864841 0.789273 0.370031 0.448256 0.178515
2001-02-28 0.991861 0.079215 0.900788 0.666178 0.693887
2001-03-31 0.016674 0.855109 0.984115 0.436574 0.480339
2001-04-30 0.120924 0.046013 0.659807 0.210534 0.694029
2001-05-31 0.788149 0.296244 0.478201 0.845042 0.437814
b 2001-01-31 0.497646 0.349958 0.223227 0.812331 0.975012
2001-02-28 0.542572 0.472267 0.276186 0.970909 0.138683
2001-03-31 0.960813 0.666942 0.069349 0.282741 0.127992
2001-04-30 0.491422 0.678742 0.048784 0.612312 0.713472
2001-05-31 0.718721 0.504403 0.069047 0.253682 0.836386
I can still slice using the method above on any specific level by:
df.loc['a']['2012-01':'2012-03']
But this is only for level0 == 'a'.
How do I do this for all values in level0? I expect something like this:
A B C D E
a 2001-01-31 0.864841 0.789273 0.370031 0.448256 0.178515
2001-02-28 0.991861 0.079215 0.900788 0.666178 0.693887
2001-03-31 0.016674 0.855109 0.984115 0.436574 0.480339
b 2001-01-31 0.497646 0.349958 0.223227 0.812331 0.975012
2001-02-28 0.542572 0.472267 0.276186 0.970909 0.138683
2001-03-31 0.960813 0.666942 0.069349 0.282741 0.127992
Use pd.IndexSlice
df.loc[pd.IndexSlice[:, '2001-01':'2001-3'], :]
来源：https://stackoverflow.com/questions/38346668/time-slice-on-second-level-of-multiindex