Pandas 数据透视表嵌套 Aggfunc

Question

我正在尝试创建一个数据透视表来计算表单的数量以及该计数的总和、平均值和中位数。但是，Forms dtype 是分类的，我不能对非数字值使用均值和中值函数。

我想使用“表单类型”：计数作为我的聚合值。

如果我将平均值包含在第一个 aggfunc 中，我会收到此错误

DataError: No numeric types to aggregate

有没有办法在数据透视表中嵌套 aggfuncts 或在创建数据透视表时更改 dtype？

我尝试过使用 .astype(int) 但我似乎无法弄清楚该函数的语法。

下面的虚拟代码并不完全准确，但我可以继续努力。

df = pd.DataFrame({"A": ["foo", "foo", "foo", "foo", "foo", "bar", "bar", "bar", "bar"],
                   "B": ["one", "one", "one", "two", "two", "one", "one", "two", "two"],
                   "C": ["105319", "1271075", "84565", "84354", "54835", "81638", "1282224", "41856", "78987"],
                   "Form Type": ["144", "D", "D/A", "144", "D", "D", "D", "S-1","D"]})

table = pd.pivot_table(df, columns = ['Form Type'],
                       index=['A', 'B', 'C'],
                       fill_value =' ', 
                       aggfunc={'Form Type': ['count']})

输出看起来与此类似

Answer 1

使用.value_counts() 计算Form Type：

• 这假设您要查询Form Type 的计数，然后您需要counts 的mean、median 和sum

import pandas as pd

df = pd.DataFrame({"A": ["foo", "foo", "foo", "foo", "foo", "bar", "bar", "bar", "bar"],
                   "B": ["one", "one", "one", "two", "two", "one", "one", "two", "two"],
                   "C": ["105319", "1271075", "84565", "84354", "54835", "81638", "1282224", "41856", "78987"],
                   "Form Type": ["144", "D", "D/A", "144", "D", "D", "D", "S-1","D"]})


counts = df['Form Type'].value_counts().rename_axis('Form Type').reset_index(name='counts')

Form Type  counts
        D       5
      144       2
      S-1       1
      D/A       1

counts['counts'].agg(['mean', 'median', 'sum'])

# Output:
mean      2.25
median    1.50
sum       9.00
Name: counts, dtype: float64

替代方案：

• 如果你想得到基于groupby的.value_counts

grouped_counts = df.groupby(['A', 'B'])['Form Type'].value_counts().rename_axis(['A', 'B', 'Form Type']).reset_index(name='counts')

   A    B Form Type  counts
 bar  one         D       2
 bar  two         D       1
 bar  two       S-1       1
 foo  one       144       1
 foo  one         D       1
 foo  one       D/A       1
 foo  two       144       1
 foo  two         D       1

grouped_counts_stats = grouped_counts.groupby(['A', 'B', 'Form Type'])['counts'].agg(['mean', 'median', 'sum'])