旋转数据框的想法：从长到宽

Question

我有一个数据框数据记录堆叠，同一主题每 3 个月左右有不同的测量值。例如Subj BAR02002有4条不同的数据记录：

    Subj  months   X    Y    Z
BAR02002   0       14  53   52
BAR02002   3       24  61   96
BAR02002   6       5   53   3
BAR02002   9       3   64   33
BAR02003   0       22  63   55
BAR02003   6       44  22   53 
BAR02003   9       42  12   72
BAR02003   12      15  1    12

我正在尝试让 BAR02002 只构成一行而不是 4。我相信这个过程被称为从长到宽的重塑数据（我可能是错的）。为了说明最终结果：

Subj       X    Y    Z    X1    Y2    Z3   X2   Y3   Z3  ... 
BAR02002   14   53   52   24    61    96   5    53    3  ...    
BAR02003   0    22   63   55    NA    NA   NA   44   22  ...

下面的代码没有给出我想要的。有什么方法可以使用 pandas/python（甚至 R）转换数据吗？

df.pivot(index='Subj_FU', columns='Subj', values= ['Months','PM_N', ...])

Answer 1

将map用于新列并将其用于参数columns，最后展平MultiIndex：

df['g'] = df['months'].map({0:0, 3:1, 6:2, 9:3, 12:4})
df1 = df.pivot_table(index='Subj', columns='g', values= ['X','Y','Z'], aggfunc='sum')
df1.columns = df1.columns.map(lambda x: f'{x[0]}{x[1]}')
print (df1)
            X0    X1    X2    X3    X4    Y0    Y1    Y2    Y3   Y4    Z0  \
Subj                                                                        
BAR02002  14.0  24.0   5.0   3.0   NaN  53.0  61.0  53.0  64.0  NaN  52.0   
BAR02003  22.0   NaN  44.0  42.0  15.0  63.0   NaN  22.0  12.0  1.0  55.0   

            Z1    Z2    Z3    Z4  
Subj                              
BAR02002  96.0   3.0  33.0   NaN  
BAR02003   NaN  53.0  72.0  12.0

如果使用列 month:

df1 = df.pivot_table(index='Subj', columns='months', values= ['X','Y','Z'], aggfunc='sum')
df1.columns = df1.columns.map(lambda x: f'{x[0]}{x[1]}')
print (df1)
            X0    X3    X6    X9   X12    Y0    Y3    Y6    Y9  Y12    Z0  \
Subj                                                                        
BAR02002  14.0  24.0   5.0   3.0   NaN  53.0  61.0  53.0  64.0  NaN  52.0   
BAR02003  22.0   NaN  44.0  42.0  15.0  63.0   NaN  22.0  12.0  1.0  55.0   

            Z3    Z6    Z9   Z12  
Subj                              
BAR02002  96.0   3.0  33.0   NaN  
BAR02003   NaN  53.0  72.0  12.0

或使用Series.unstack:

g = df['months'].map({0:0, 3:1, 6:2, 9:3, 12:4})
df1 = df.groupby(['Subj', g])[['X','Y','Z']].sum().unstack()
df1.columns = df1.columns.map(lambda x: f'{x[0]}{x[1]}')

Answer 2

您可以简单地 drop 重复项，它会保留第一项：

import pandas as pd

data = [ { "Subj": "BAR02002", "months": 0, "X": 14, "Y": 53, "Z": 52 }, { "Subj": "BAR02002", "months": 3, "X": 24, "Y": 61, "Z": 96 }, { "Subj": "BAR02002", "months": 6, "X": 5, "Y": 53, "Z": 3 }, { "Subj": "BAR02002", "months": 9, "X": 3, "Y": 64, "Z": 33 }, { "Subj": "BAR02003", "months": 0, "X": 22, "Y": 63, "Z": 55 }, { "Subj": "BAR02003", "months": 6, "X": 44, "Y": 22, "Z": 53 }, { "Subj": "BAR02003", "months": 9, "X": 42, "Y": 12, "Z": 72 }, { "Subj": "BAR02003", "months": 12, "X": 15, "Y": 1, "Z": 12 } ]
df = pd.DataFrame(data)

结果：

	Subj	months	X	Y	Z
0	BAR02002	0	14	53	52
4	BAR02003	0	22	63	55

旋转数据框的想法：从长到宽

Pivoting dataframe ideas: from long to wide

python

r

pandas

pivot-table

reshape