pandas pivot_table 的边距仅垂直聚合 [英] pandas pivot_table's margins is only aggregating vertically

问题描述

考虑一个数据框:

df = pd.DataFrame(
    {'last_year': [1, 2, 3], 'next_year': [4, 5, 6]}, 
    index=['foo', 'bar', 'star']
)

      last_year  next_year
foo           1          4
bar           2          5
star          3          6

我正在寻找一种简单的方法来显示此表周围的总计，包括列和行.

我的想法是通过 .pivot_table() 抛出它:

My thought was to throw it through .pivot_table():

pd.pivot_table(
    df,
    index=df.index,
    margins=True,
    aggfunc=sum
)

然而，这只适用于第一个轴(垂直):

However, this only works for the first axis (vertically):

      last_year  next_year
bar           2          5
foo           1          4
star          3          6
All           6         15

我错过了什么?为什么不计算行明智的总数，就像在 本文档的示例?另外，为什么会打乱我的索引顺序?

What did I miss? How come no row wise totals are calculated as well, just like in this documentation's example? Also, why is it messing up my index' order?

我对 df['All'] = df.sum(axis=1) 类型的解决方案不感兴趣；我想要一种不影响我原始数据框的动态方法.数据透视表似乎是最合乎逻辑的方法(据我所知)，但也许有更好的方法！

I'm not interested in df['All'] = df.sum(axis=1) kind of solutions; I want a dynamic approach which does not affect my original dataframe. A pivot table seems like the most logical way to do this (as far as I know) but maybe there are better ways!

推荐答案

我的猜测是 pivot_table 中的每一列都是它自己的组，所以你看不到水平聚合(这是毫无意义的聚合一个元素).为了演示差异，请考虑 stack():

My guess is that each column is its own group in pivot_table so you don't see the horizontal aggregation (it's rather pointless to aggregate one single element). To demonstrate the difference, consider stack():

(df.stack().reset_index(name='value')
   .pivot_table(index='level_0', columns='level_1', values='value', margins=True,
               aggfunc='sum')
) 

输出:

level_1  last_year  next_year  All
level_0                           
bar              2          5    7
foo              1          4    5
star             3          6    9
All              6         15   21

也就是说，这很可能是一个错误.

That said, it might very well be a bug.

这篇关于pandas pivot_table 的边距仅垂直聚合的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！