如果您想要seaborn.lineplot 提供的错误带/条,您必须自己绘制它们。下面是几个示例,说明如何在 matplotlib 中绘制误差带和误差条,并获得与 seaborn 中相似的绘图。它们是使用作为 pandas 数据框导入的 fmri 示例数据集构建的,并且基于 lineplot function 上的 seaborn 文档中显示的示例之一。
import numpy as np # v 1.19.2
import pandas as pd # v 1.1.3
import matplotlib.pyplot as plt # v 3.3.2
import seaborn as sns # v 0.11.0
# Import dataset as a pandas dataframe
df = sns.load_dataset('fmri')
# display(df.head(3))
subject timepoint event region signal
0 s13 18 stim parietal -0.017552
1 s5 14 stim parietal -0.080883
2 s12 18 stim parietal -0.081033
此数据集包含一个名为 timepoint 的时间变量,在 19 个时间点中的每一个时间点对 信号 进行了 56 次测量。我使用默认估计器,即平均值。为了简单起见,我没有使用平均值的标准误差的置信区间作为不确定性(又名误差)的度量,而是使用每个时间点的测量值的标准偏差。这是通过传递ci='sd' 在lineplot 中设置的,误差扩展到平均值每一侧的一个标准差(即对称)。下面是带有误差带的 seaborn 线图(默认情况下):
# Draw seaborn lineplot with error band based on the standard deviation
fig, ax = plt.subplots(figsize=(9,5))
sns.lineplot(data=df, x="timepoint", y="signal", ci='sd')
sns.despine()
plt.show()
现在假设我更希望有一个误差带,该误差带跨越平均值两侧每个时间点测量值的标准差的一半。由于在调用 lineplot 函数时无法设置此首选项,据我所知,最简单的解决方案是使用 matplotlib 从头开始创建绘图。
# Matplotlib plot with custom error band
# Define variables to plot
y_mean = df.groupby('timepoint').mean()['signal']
x = y_mean.index
# Compute upper and lower bounds using chosen uncertainty measure: here
# it is a fraction of the standard deviation of measurements at each
# time point based on the unbiased sample variance
y_std = df.groupby('timepoint').std()['signal']
error = 0.5*y_std
lower = y_mean - error
upper = y_mean + error
# Draw plot with error band and extra formatting to match seaborn style
fig, ax = plt.subplots(figsize=(9,5))
ax.plot(x, y_mean, label='signal mean')
ax.plot(x, lower, color='tab:blue', alpha=0.1)
ax.plot(x, upper, color='tab:blue', alpha=0.1)
ax.fill_between(x, lower, upper, alpha=0.2)
ax.set_xlabel('timepoint')
ax.set_ylabel('signal')
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.show()
如果您更喜欢有误差线,这就是 seaborn 线图的样子:
# Draw seaborn lineplot with error bars based on the standard deviation
fig, ax = plt.subplots(figsize=(9,5))
sns.lineplot(data=df, x="timepoint", y="signal", ci='sd', err_style='bars')
sns.despine()
plt.show()
以下是如何使用自定义误差线获得与 matplotlib 相同类型的绘图:
# Matplotlib plot with custom error bars
# If for some reason you only have lists of the lower and upper bounds
# and not a list of the errors for each point, this seaborn function can
# come in handy:
# error = sns.utils.ci_to_errsize((lower, upper), y_mean)
# Draw plot with error bars and extra formatting to match seaborn style
fig, ax = plt.subplots(figsize=(9,5))
ax.errorbar(x, y_mean, error, color='tab:blue', ecolor='tab:blue')
ax.set_xlabel('timepoint')
ax.set_ylabel('signal')
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.show()
# Note: in this example, y_mean and error are stored as pandas series
# so the same plot can be obtained using this pandas plotting function:
# y_mean.plot(yerr=error)
Matplotlib 文档:fill_between、specify error bars、subsample error bars
熊猫文档:error bars