如何避免在网格图中使用 seaborn 进行自动排序？ （Python）

Question

我正在尝试绘制一些我拥有的数据，并且我正在努力通过 seaborn 进行自动排序。正如您在下图中看到的那样，每个图表的 x 轴排序都不同，我希望它们相同，[Melhorou, Piorou, Indiferente]。有谁知道如何做到这一点？

这是我到目前为止所做的一段代码。

f = plt.figure()

ax = f.add_subplot(2, 2, 1)
sns.despine(f)
sns.histplot(data = data , x = "A pandemia afetou a sua performance acadêmica?  [1o sem de 2020]", multiple="dodge", palette = "Blues", ax = ax);
plt.xlabel("1o sem 2020")

ax = f.add_subplot(2, 2, 2)
sns.despine(f)
sns.histplot(data = data , x = "A pandemia afetou a sua performance acadêmica?  [2o sem de 2020]", multiple="dodge", palette = "Blues", ax = ax);
plt.xlabel("2o sem 2020")

ax = f.add_subplot(2, 2, 3)
sns.despine(f)
sns.histplot(data = data , x = "A pandemia afetou a sua performance acadêmica?  [1o sem de 2021]", multiple="dodge", palette = "Blues", ax = ax);
plt.xlabel("1o sem 2021")

ax = f.add_subplot(2, 2, 4)
sns.despine(f)
sns.histplot(data = data , x = "A pandemia afetou a sua performance acadêmica?  [2o sem de 2021]", multiple="dodge", palette = "Blues", ax = ax);
plt.xlabel("2o sem 2021")

Answer 1

默认情况下，值在列中出现的顺序决定了histplot的x轴顺序。

您可以将数据框的列设为分类，并以这种方式强制排序。

from matplotlib import pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np

column_names_long = ["A pandemia afetou a sua performance acadêmica?  [1o sem de 2020]",
                     "A pandemia afetou a sua performance acadêmica?  [2o sem de 2020]",
                     "A pandemia afetou a sua performance acadêmica?  [1o sem de 2021]",
                     "A pandemia afetou a sua performance acadêmica?  [2o sem de 2021]"]
column_names_short = ["1o sem de 2020", "2o sem de 2020", "1o sem de 2021", "2o sem de 2021"]

affectings = ['Melhorou', 'Piorou', 'Indiferente']

data = pd.DataFrame({c: np.random.choice(affectings, 100) for c in column_names_long})
# rename the columns, so seaborn can show them directly
data = data.rename(columns={long: short for long, short in zip(column_names_long, column_names_short)})
for column_name in column_names_short:
    data[column_name] = pd.Categorical(data[column_name], affectings)  # fix an order on each of the columns

fig, axs = plt.subplots(ncols=2, nrows=2, figsize=(12, 7))

for ax, column_name in zip(axs.flatten(), column_names_short):
    sns.histplot(data=data, x=column_name, color="dodgerblue", ax=ax)

sns.despine(fig)

plt.tight_layout()  # fit the subplots and their labels nicely
plt.show()

或者，您可以将histplot 替换为countplot，它接受order= 参数并允许您为条形分配不同的颜色：

for ax, column_name in zip(axs.flatten(), column_names_short):
    sns.countplot(data=data, x=column_name, order=affectings,
                  palette=["dodgerblue", "crimson", "orange"], ax=ax)