pandas 数据透视表：如何在索引和列中查找每个组的计数

Question

I am new to pivot table. I just want the count of each Age Grp by Mth. I tried

pivot1=pd.pivot_table(df,index=[ "Age Grp"], columns=["Mth"], values=["Age Grp"], aggfunc=pd.Series.nunique)

but get ValueError: Grouper for 'Age Grp' not 1-dimensional

DF

 |Age Grp | Mth  | 
 | 0-4    |  1   |  
 | 5 -9   | 5    |  
 | 0-4    | 10   | 
 | 10-14  | 5  |  

Desired Outcome

     Mth  |  1    | 5 | 10 |
 |Age Grp |                |
 | 0-4    |  1   |   0  | 1|
 | 5 -9   | 0    |   1  | 0|
 | 10-14  |  0   |   1  | 0|

Answer 1

We can try crosstab instead of pivot_table with values and an aggfunc then use fillna to replace the NaN values with 0 :

ct_df = pd.crosstab(
    df['Age Grp'], df['Mth'],
    values=df['Age Grp'], aggfunc='nunique'
).fillna(0, downcast='infer')

ct_df :

Mth      1   5   10
Age Grp            
0-4       1   0   1
10-14     0   1   0
5-9       0   1   0

Or with groupby nunique + unstack :

df_us = df.groupby(
    ['Age Grp', 'Mth']
)['Age Grp'].nunique().unstack(fill_value=0)

df_us :

Mth      1   5   10
Age Grp            
0-4       1   0   1
10-14     0   1   0
5-9       0   1   0

---

Setup and imports

import pandas as pd  # 1.3.4

df = pd.DataFrame({
    'Age Grp': ['0-4', '5-9', '0-4', '10-14'], 'Mth': [1, 5, 10, 5]
})

Answer 2

Or:

df.groupby(['Age Grp', 'Mth']).size().unstack(fill_value=0)

Output:

Mth      1   5   10
Age Grp            
0-4       1   0   1
10-14     0   1   0
5-9       0   1   0