参考一：https://www.cnblogs.com/laoketeng/p/11268581.html

library(plm)

library(psych)

library(xts)

library(tseries)

library(lmtest)

 

## import dataset

datas<-read.table("data.txt",header =TRUE)

 

## adf test

pcgdp<-xts(datas$PCGDP,as.Date(datas$year))

adf.test(pcgdp)

# result： stationary

 

ltax<-xts(datas$Ltax,as.Date(datas$year))

adf.test(ltax)

# result： stationary

 

hp<-xts(datas$hp,as.Date(datas$year))

adf.test(hp)

# result： stationary

 

lp<-xts(datas$lp,as.Date(datas$year))

adf.test(lp)

# result： stationary

 

## 协整检验

# Engle-Granger

reg<-lm(datas$hp~datas$lp+datas$Ltax+datas$PCGDP)

summary(reg)

error<-residuals(reg)

adf.test(error)

# result: residuals stationary

 

### 面板数据回归

hpdatas<-plm.data(datas,index=c("city","year"))

 

# Pooled Regression Model

hp_pool<-plm(hp~lp+Ltax+PCGDP+PP,data=hpdatas,model = "pooling")

 

# Fixed Effects Regression Model

hp_fe<-plm(hp~lp+Ltax+PCGDP+PP,data=hpdatas,model = "within")

 

# F-test :

pFtest(hp_fe,hp_pool)

# result: significant effects

 

# Random Effects Regression Model


hp_re<-plm(hp~lp+Ltax+PCGDP,data=hpdatas,model="random",random.method = "swar")

           

# Hausman test

phtest(hp_fe,hp_re)

# if p<0.05,then use fixed effects

# result: p=0.6785>0.05,use random ffects


# Random Effects Regression Model

hp_re<-plm(hp~lp+Ltax+PCGDP,data=hpdatas,model="random",random.method = "swar")

summary(hp_re)

# 显著水平 a=0.01

# result: fp:房价与 lp:地价正相关，且显著； 

#         fp:房价与 Ltax: 地税收入正相关，且显著； 

#         fp:房价与 PCGDP: 人均GDP 正相关，且显著；

参考二：https://zhuanlan.zhihu.com/p/24877529

Panel Data Models in R

library(plm)
mydata <- read.csv("panel_wage.csv")
attach(mydata)
Y <- cbind(lwage)
X <- cbind(exp, exp2, wks, ed)

summary(Y)
summary(X)

ols <- lm(Y ~ X)
summary(ols)

声明面板

Stata: xtset

pdata <- plm.data(mydata,indexes = c("id","t"))

混合回归

Stata: reg

pooling <- plm(Y ~ X, data = pdata,model = "pooling")
summary(pooling)

固定效应

Stata: xtreg ,fe

fixed <- plm(Y ~ X,data = pdata,model = "within")
summary(fixed)

随机效应

Stata: xtreg ,re

random <- plm(Y ~ X,data = pdata,model = "random")
summary(random)

不同模型的比较

random vs ols

plmtest(pooling)

fixed vs ols

pFtest(fixed, pooling)

random vs fixed

phtest(random, fixed)