【定量分析、量化金融与统计学】R语言：哑变量与多重共线性

一、前言

在【定量分析、量化金融与统计学】多元回归模型与回归推理之中

我们对多重共线性做了简单解释，今天用程序来做一下多重共线性的检验。

二、程序分析

加载数据到数据帧并显示数据内容：

x<-read.csv(file="C:/R_files/Employee_salary.csv", header=T)
head(x)
names(x)

画散点图简单观察情况：

plot(x$experience,x$salary, xlab="Experience", ylab="Salary", main="Scatter Plot")

创建空列 

dummyGender<-NULL
dummyGender<-(x$gender=="female")*1

将其添加到数据集中

x1<-cbind(dummyGender, x)
head(x1)

R会自己生成哑变量

做线性模型：

fit1<-lm(salary~experience+as.factor(gender), x)
summary(fit1)

 测试交互效应：

fit1<-lm(salary~experience+as.factor(gender)+experience:as.factor(gender), x)
summary(fit1)

 测试年龄以及年龄与经验（不交互）的结果：

fit2<-lm(salary~age+as.factor(gender), x)
summary(fit2)

fit3<-lm(salary~experience+age+as.factor(gender), x)
summary(fit3)

 相关矩阵：

#correlation matrix
#For a better visualization you can use the mentioned package
install.packages("PerformanceAnalytics")
install.packages("zoo")
install.packages("xts")
library("zoo")
library("xts")
library("PerformanceAnalytics")
my_data <- x1[, c(1,3,4,5)]
chart.Correlation(my_data, histogram=TRUE, pch= 19)
cor(my_data)

多重共线性：

#multicolinearity
install.packages("mctest")
library(mctest)
my_data1 <- x1[, c(1,3,5)]
# Individual Multicollinearity Diagnostic Measures
#Computes different measures of multicollinearity diagnostics such as 
# TOL,variance Inflation factor (VIF),Corrected VIF (CVIF), Klein's rule

imcdiag(x = my_data1, y = x1$salary)
#try imcdiag() function with model which has 3 independent variables which is "fit3"
# this below code works instead of the earlier imcdiag()
imcdiag(fit3) 
?imcdiag