熊猫：具有agg值的groupby分位数

Question

I'm trying to group numerical values by quantiles and create columns for the sum of the values falling into the quantile bands. Here's a simplified, reproducible example:

    raw_data = {'female': [0, 1, 0, 1, 0, 1, 0, 1],
        'male': [1, 0, 1, 0, 1, 0, 1, 0],
        'number': [25000, 34000, 48600, 22000, 50000, 21000, 29000, 36000]}
df = pd.DataFrame(raw_data, columns = ['female', 'male', 'number'])
df

    female  male    number
  0 0       1       25000
  1 1       0       34000
  2 0       1       48600
  3 1       0       22000
  4 0       1       50000
  5 1       0       21000
  6 0       1       29000
  7 1       0       36000

Essentially I'm trying to achieve this:

pd.DataFrame(df['number'].quantile([.1, .2, .3, .4, .5]))

    number
0.1 21700
0.2 23200
0.3 25400
0.4 28200
0.5 31500

But in this dataframe show two new columns. One for the sum of males who's number falls into the corresponding quantile band and one for the sum of females.

Initially I thought this would be a groupby with .quantile([values]) appended, and then .agg({'male': 'sum', 'female':'sum'}) This doesn't work though. Can what I'm trying to achieve even be done?

Answer 1

You want to use pd.qcut to create the groupings:

qs = pd.qcut(df.number, [0, .1, .2, .3, .4, .5, 1], ['q%d' % i for i in xrange(6)])
qs

0    q2
1    q5
2    q5
3    q1
4    q5
5    q0
6    q4
7    q5
Name: number, dtype: category
Categories (6, object): [q0 < q1 < q2 < q3 < q4 < q5]

Then groupby and agg

df.groupby(qs).agg({'male': 'sum', 'female':'sum'})