从列表列中使用seaborn进行分组箱线图

Question

我在 python panda DataFrame 中有以下数据。我想要类似于 https://stanford.edu/~mwaskom/software/seaborn/examples/grouped_boxplot.html

中的分组箱形图

对于每个 id，我希望并排绘制两个箱形图。我该如何实现这一目标。我尝试用 seaborn 包绘制它，但没有成功。

id               predicted                              real
1            [10, 10, 10]                      [16, 18, 20]   
2            [12, 12, 15]              [15, 17, 19, 21, 23]
3           [20, 5, 4, 4]                          [29, 32]
4    [25, 25, 25, 24, 21]  [21, 24, 25, 26, 28, 29, 30, 33]
5        [20, 20, 20, 21]  [21, 22, 24, 26, 28, 30, 31, 32]
6           [8, 3, 3, 14]                          [25, 27]
7  [1, 4, 4, 4, 5, 6, 10]                      [69, 71, 72]
8        [11, 11, 11, 11]              [19, 21, 22, 23, 24]
9            [7, 6, 9, 9]                  [19, 26, 27, 28]
10    [30, 30, 30, 30, 30]                          [38, 39]

Answer 1

注意您正在查看的示例中的表结构

import seaborn as sns

tips = sns.load_dataset("tips")
sns.boxplot(x="day", y="total_bill", hue="sex", data=tips, palette="PRGn")
sns.despine(offset=10, trim=True)

tips.head()

---

我们的目标是将您的桌子布置成这样

设置

from StringIO import StringIO
import pandas as pd

text = """id               predicted                              real
1            [10, 10, 10]                      [16, 18, 20]   
2            [12, 12, 15]              [15, 17, 19, 21, 23]
3           [20, 5, 4, 4]                          [29, 32]
4    [25, 25, 25, 24, 21]  [21, 24, 25, 26, 28, 29, 30, 33]
5        [20, 20, 20, 21]  [21, 22, 24, 26, 28, 30, 31, 32]
6           [8, 3, 3, 14]                          [25, 27]
7  [1, 4, 4, 4, 5, 6, 10]                      [69, 71, 72]
8        [11, 11, 11, 11]              [19, 21, 22, 23, 24]
9            [7, 6, 9, 9]                  [19, 26, 27, 28]
10    [30, 30, 30, 30, 30]                          [38, 39]"""

df = pd.read_csv(StringIO(text), sep='\s{2,}', engine='python', index_col=0)

df = df.stack().str.strip('[]') \
       .str.split(', ').unstack()
df

---

以正确的格式获取表格

df1 = df.stack().apply(pd.Series).stack().astype(int) \
          .rename_axis(['id', 'reality', None]) \
          .rename('value').reset_index(['id', 'reality']) \
          .reset_index(drop=True)

df1.head()

sns.boxplot(x='id', y='value', hue='reality', data=df1, palette='PRGn')
sns.despine(offset=10, trim=True)

Answer 2

• 给定 list type 的

  pandas

  列，最简单的方法是将

  pandas.DataFrame.melt

  列转换为长格式，然后将

  pandas.DataFrame.explode

  列表转换为单独的值。

• sns.catplot

  与

  kind='box'

  创建一个

  figure-level

  图，而这个 answer 创建一个

  axes-level

  图。
  • 图形级函数与轴级函数

import pandas as pd
import seaborn as sns

# convert df to long-form and explode the lists
dfm = df.melt(id_vars='id').explode('value')

# plot figure-level boxplot
g = sns.catplot(kind='box', data=dfm, x='id', y='value', hue='variable', height=5, aspect=2)

df

data = {'id': [1, 2, 3, 4, 5, 6, 7, 8, 9, 10], 'predicted': [[10, 10, 10], [12, 12, 15], [20, 5, 4, 4], [25, 25, 25, 24, 21], [20, 20, 20, 21], [8, 3, 3, 14], [1, 4, 4, 4, 5, 6, 10], [11, 11, 11, 11], [7, 6, 9, 9], [30, 30, 30, 30, 30]],
        'real': [[16, 18, 20], [15, 17, 19, 21, 23], [29, 32], [21, 24, 25, 26, 28, 29, 30, 33], [21, 22, 24, 26, 28, 30, 31, 32], [25, 27], [69, 71, 72], [19, 21, 22, 23, 24], [19, 26, 27, 28], [38, 39]]}

df = pd.DataFrame(data)

dfm

    id   variable value
0    1  predicted    10
0    1  predicted    10
0    1  predicted    10
1    2  predicted    12
1    2  predicted    12
1    2  predicted    15
2    3  predicted    20
2    3  predicted     5
2    3  predicted     4
2    3  predicted     4
3    4  predicted    25
3    4  predicted    25
3    4  predicted    25
3    4  predicted    24
3    4  predicted    21
4    5  predicted    20
4    5  predicted    20
4    5  predicted    20
4    5  predicted    21
5    6  predicted     8
5    6  predicted     3
5    6  predicted     3
5    6  predicted    14
6    7  predicted     1
6    7  predicted     4
6    7  predicted     4
6    7  predicted     4
6    7  predicted     5
6    7  predicted     6
6    7  predicted    10
7    8  predicted    11
7    8  predicted    11
7    8  predicted    11
7    8  predicted    11
8    9  predicted     7
8    9  predicted     6
8    9  predicted     9
8    9  predicted     9
9   10  predicted    30
9   10  predicted    30
9   10  predicted    30
9   10  predicted    30
9   10  predicted    30
10   1       real    16
10   1       real    18
10   1       real    20
11   2       real    15
11   2       real    17
11   2       real    19
11   2       real    21
11   2       real    23
12   3       real    29
12   3       real    32
13   4       real    21
13   4       real    24
13   4       real    25
13   4       real    26
13   4       real    28
13   4       real    29
13   4       real    30
13   4       real    33
14   5       real    21
14   5       real    22
14   5       real    24
14   5       real    26
14   5       real    28
14   5       real    30
14   5       real    31
14   5       real    32
15   6       real    25
15   6       real    27
16   7       real    69
16   7       real    71
16   7       real    72
17   8       real    19
17   8       real    21
17   8       real    22
17   8       real    23
17   8       real    24
18   9       real    19
18   9       real    26
18   9       real    27
18   9       real    28
19  10       real    38
19  10       real    39