No method to rename Pandas index names (not labels) inline?-CodePudding

I have a dataframe bmi. I can summarize it like this:

Now I want to summarize the dataframe based on Age values measured in decades. So I do the following:

bmi.groupby(by=bmi.Age//10).describe().stack()

which shows me the following summary:

Note that this creates an index named Age which I'd rather name Decade. How can I rename the name of this index (to be precise, index level) during the generation of the dataframe?

So I am looking for a function which_func() so that my code can read like this:

bmi.groupby(by=bmi.Age//10).describe().which_func({'Age':'Decade'}).stack()

and I get the output:

The shortest single-line version I can come up with is:

bmi.groupby(by=bmi.Age//10).describe().reset_index(level='Age', names='Decades').set_index('Decades').stack()

which seems redundant.

The built-in method DataFrame.rename() renames the labels of an index, not its name, and other methods require extracting the index or their names and then renaming, overwriting them so the code cannot be chained.

Given that this is a very common situation in groupby()-aggregate() setup, is there really no way to do this with chained code?

CodePudding user response：

What you are looking for is rename_axis():

>>> bmi.groupby(by=bmi.Age//10).describe().rename_axis(index={'Age': 'Decade'}).stack()
                Bodyfat        Age     Weight
Decade                                       
1.0    count   9.000000   9.000000   9.000000
       mean   30.779899  16.981942  32.795294
       std     4.341956   1.415877   5.292009
       min    24.080597  14.769285  24.887466
       25%    27.655679  16.137036  30.101671
...                 ...        ...        ...
4.0    min    27.594864  41.519583  24.699921
       25%    27.594864  41.519583  24.699921
       50%    27.594864  41.519583  24.699921
       75%    27.594864  41.519583  24.699921
       max    27.594864  41.519583  24.699921

For the actual result above, we used the following reproducible example:

# minimal example
import numpy as np  # only needed to create the example frame

np.random.seed(0)
bmi = pd.DataFrame(np.random.normal(30, 5, size=(250, 3)),
                   columns='Bodyfat Age Weight'.split())

The following forms are all equivalent:

bmi.groupby(by=bmi.Age//10).describe().rename_axis(index='Decade').stack()

bmi.groupby(by=bmi.Age//10).describe().rename_axis('Decade', axis=0).stack()

bmi.groupby(by=bmi.Age//10).describe().stack().rename_axis(index={'Age': 'Decade'})

(the last one shows that rename_axis() can be called after .stack(), meaning it can be used on a MultiIndex as well).

You can of course also derive a new variable Decade before groupby:

>>> bmi.assign(Decade=bmi['Age']//10).groupby('Decade').describe().stack()
                Bodyfat        Age     Weight
Decade                                       
1.0    count   9.000000   9.000000   9.000000
       mean   30.779899  16.981942  32.795294
       std     4.341956   1.415877   5.292009
       min    24.080597  14.769285  24.887466
       25%    27.655679  16.137036  30.101671
...                 ...        ...        ...
4.0    min    27.594864  41.519583  24.699921
       25%    27.594864  41.519583  24.699921
       50%    27.594864  41.519583  24.699921
       75%    27.594864  41.519583  24.699921
       max    27.594864  41.519583  24.699921