如何绘制标准偏差

How to plot Standard Deviations

最近我开始学习数据科学的概率和统计。我正在尝试为以下分布 X 绘制 标准差 ,例如 68-95-99.7 规则。

生成绘图的代码:

import numpy as np
from scipy.stats import norm

import matplotlib.pyplot as plt
import seaborn as sns

# Line width: Maximum 130 characters in the output, post which it will continue in next line.
np.set_printoptions(linewidth=130)

sns.set_context("paper", font_scale=1.5)

# Distribution
X = [1, 2, 2, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 5, 6, 6, 6, 6, 7, 7, 7, 8, 8, 9]

mean = np.mean(X)
var = np.var(X)
std = np.std(X)

print("Mean:", mean)
print("Variance:", var)
print("Standard Deviation:", std)

"""
Mean: 5.0
Variance: 4.0
Standard Deviation: 2.0
"""

plt.figure(figsize=(10, 5))

ax = sns.kdeplot(X, shade=True)

# Plot 1-std
x = np.linspace(mean - std, mean + std)
y = norm.pdf(x, mean, std)
ax.fill_between(x, y, alpha=0.5)

plt.xlabel("Random variable X")
plt.ylabel("Probability Density Function")
plt.xticks(ticks=range(0, 10))
plt.grid()

plt.show()

此代码生成以下图:

问题:

  1. 根据均值绘制 1 个标准差的代码有什么问题?
  2. 我无法理解为什么 kde 图上方有一个小峰?
  3. 如何绘制 1-std、2-std 和 3-std?
  1. 您的代码没有错:mean 是 5 和 std 2,因此您正在遮蔽 5 - 2 = 35 + 2 = 7 之间的区域。

  2. kde 图中有一个小峰,因为它表示您用 X 给出的数据分布,实际上,X 是不是正态分布。您可以使用真正的正态分布来检查这一点:

    mean = 5
    std = 2
    X = np.random.randn(10000)
    X = (X - X.mean())/X.std()*std + mean
    

  3. 您可以在 i 上使用 for 循环绘制其他标准偏差。 x1 是左侧,x2 是中间部分(然后设置为 np.nan),最后 x3 是分布的右侧。然后你必须设置 np.nan 要排除的区域(对应于 x2):

    N = 10
    for i in [1, 2, 3]:
        x1 = np.linspace(mean - i*std, mean - (i - 1)*std, N)
        x2 = np.linspace(mean - (i - 1)*std, mean + (i - 1)*std, N)
        x3 = np.linspace(mean + (i - 1)*std, mean + i*std, N)
        x = np.concatenate((x1, x2, x3))
        x = np.where((mean - (i - 1)*std < x) & (x < mean + (i - 1)*std), np.nan, x)
        y = norm.pdf(x, mean, std)
        ax.fill_between(x, y, alpha=0.5)
    

完整代码

import numpy as np
from scipy.stats import norm

import matplotlib.pyplot as plt
import seaborn as sns

# Line width: Maximum 130 characters in the output, post which it will continue in next line.
np.set_printoptions(linewidth=130)

sns.set_context("paper", font_scale=1.5)

# Distribution
X = [1, 2, 2, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 5, 6, 6, 6, 6, 7, 7, 7, 8, 8, 9]


mean = np.mean(X)
var = np.var(X)
std = np.std(X)

print("Mean:", mean)
print("Variance:", var)
print("Standard Deviation:", std)

"""
Mean: 5.0
Variance: 4.0
Standard Deviation: 2.0
"""

plt.figure(figsize=(10, 5))

ax = sns.kdeplot(X, shade=True)

N = 10
for i in [1, 2, 3]:
    x1 = np.linspace(mean - i*std, mean - (i - 1)*std, N)
    x2 = np.linspace(mean - (i - 1)*std, mean + (i - 1)*std, N)
    x3 = np.linspace(mean + (i - 1)*std, mean + i*std, N)
    x = np.concatenate((x1, x2, x3))
    x = np.where((mean - (i - 1)*std < x) & (x < mean + (i - 1)*std), np.nan, x)
    y = norm.pdf(x, mean, std)
    ax.fill_between(x, y, alpha=0.5)

plt.xlabel("Random variable X")
plt.ylabel("Probability Density Function")
plt.xticks(ticks=range(0, 10))
plt.grid()

plt.show()

情节