用python计算残差等

一、简介

 这里简单介绍了回归诊断，本文简单介绍如何用python计算其中的值。

二、计算

import statsmodels.api as sm
# 以波士顿房价为例
from sklearn.datasets import load_boston
from sklearn.model_selection import train_test_split

X = load_boston()['data']
y = load_boston()['target']

# 加上全1列
X = sm.add_constant(X)

# 划分训练集和测试集
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=2021)

# 建立线性回归模型
ols = sm.OLS(y_train, X_train)
models = ols.fit()

# 计算预测值
y_predict = models.predict(X_train)
outliers = models.get_influence()

2.1 计算残差

y_train - y_predict # (n_samples,)

2.2 计算学生化残差

resids1 = outliers.resid_studentized_external  # (n_samples,)

或

resids2 = outliers.resid_studentized_internal # (n_samples,)

我也没搞懂这两个有啥区别，它们之间的数值差的比较小。

2.3 画残差图

plt.scatter(y_predict, resids1)
plt.xlabel('y_predict')
plt.ylabel('resid')
plt.yticks(range(-5, 6)) 
plt.axhline(y=2, color='r', linestyle='--')
plt.axhline(y=-2, color='r', linestyle='--') 
plt.show()

2.4 计算Cook距离

cook = outliers.cooks_distance # (n_samples,)

2.5 帽子矩阵

h = outliers.hat_matrix_diag # (n_samples,)

2.6 dffits值

diffts = outliers.dffits
diffts[0] # (n_samples,)
diffts[1] # (), 就一个数