时间序列分析

kt与gt的协整分析

library(urca)
s<-ur.df(k_likelihood, type = c("trend"), lags = 1,
         selectlags = c("AIC"))
summary(s)
############################################### 
# Augmented Dickey-Fuller Test Unit Root Test # 
############################################### 

Test regression trend 


Call:
lm(formula = z.diff ~ z.lag.1 + 1 + tt + z.diff.lag)

Residuals:
    Min      1Q  Median      3Q     Max 
-3.6683 -0.8258  0.0533  0.9626  3.4472 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)
(Intercept)  2.01652    2.32422   0.868    0.389
z.lag.1     -0.08669    0.05406  -1.604    0.114
tt          -0.10658    0.07113  -1.498    0.140
z.diff.lag  -0.03778    0.13163  -0.287    0.775

Residual standard error: 1.574 on 57 degrees of freedom
Multiple R-squared:  0.05296,	Adjusted R-squared:  0.003116 
F-statistic: 1.063 on 3 and 57 DF,  p-value: 0.3722


Value of test-statistic is: -1.6037 10.8166 1.4228 

Critical values for test statistics: 
      1pct  5pct 10pct
tau3 -4.04 -3.45 -3.15
phi2  6.50  4.88  4.16
phi3  8.73  6.49  5.47
##一阶的t值为-1.604>-4.04,>-3.45,>-3.15
不在拒绝域里面，接受有单位跟的原假设。
hh<-ur.df(diff(k_likelihood), type = c("none"), lags = 1,
          selectlags = c("AIC"))
summary(hh)
############################################### 
# Augmented Dickey-Fuller Test Unit Root Test # 
############################################### 

Test regression none 


Call:
lm(formula = z.diff ~ z.lag.1 - 1 + z.diff.lag)

Residuals:
    Min      1Q  Median      3Q     Max 
-3.5749 -1.8058 -0.7448  1.0218  3.2996 

Coefficients:
           Estimate Std. Error t value Pr(>|t|)    
z.lag.1     -0.3637     0.1325  -2.746 0.008031 ** 
z.diff.lag  -0.4226     0.1187  -3.561 0.000745 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 1.762 on 58 degrees of freedom
Multiple R-squared:  0.4374,	Adjusted R-squared:  0.418 
F-statistic: 22.55 on 2 and 58 DF,  p-value: 5.701e-08


Value of test-statistic is: -2.7455 

Critical values for test statistics: 
     1pct  5pct 10pct
tau1 -2.6 -1.95 -1.61
t值是-2.7455，<-2.6,<-1.95,<-1.61
在拒绝域里面，是平稳的。

> pp.test(g)

	Phillips-Perron Unit Root Test

data:  g
Dickey-Fuller Z(alpha) = -7.527, Truncation lag parameter = 3, p-value = 0.6673
alternative hypothesis: stationary
p大,接受原假设，不平稳

> pp.test(diff(g))

	Phillips-Perron Unit Root Test

data:  diff(g)
Dickey-Fuller Z(alpha) = -36.495, Truncation lag parameter = 3, p-value = 0.01
alternative hypothesis: stationary

Warning message:
In pp.test(diff(g)) : p-value smaller than printed p-value
p小，拒绝原假设，平稳

kt和gt都是不平稳的时间序列，但一阶差分平稳

求相关系数

ss<-data.frame(k_likelihood[11:63],1/g)
cor(ss)
                   k_likelihood.11.63.      X1.g
k_likelihood.11.63.           1.0000000 0.9728327
X1.g                          0.9728327 1.0000000
cor.test(k_likelihood[11:63],1/g,conf.level=0.99)
Pearson's product-moment correlation

data:  k_likelihood[11:63] and 1/g
t = 30.009, df = 51, p-value < 2.2e-16
alternative hypothesis: true correlation is not equal to 0
99 percent confidence interval:
 0.9445150 0.9867961
sample estimates:
      cor 
0.9728327 

有很强的相关性

协整分析

协整回归

> s<-lm(k_likelihood[11:63]~g)
> summary(s)

Call:
lm(formula = k_likelihood[11:63] ~ g)

Residuals:
    Min      1Q  Median      3Q     Max 
-10.241  -4.408  -0.130   4.019   8.631 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)   633.95      21.80   29.09   <2e-16 ***
g             -61.74       2.10  -29.41   <2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 4.799 on 51 degrees of freedom
Multiple R-squared:  0.9443,	Adjusted R-squared:  0.9432 
F-statistic: 864.7 on 1 and 51 DF,  p-value: < 2.2e-16

有很强的拟合优度

DW检验

> library(lmtest)
> dwtest(s)

	Durbin-Watson test

data:  s
DW = 0.23248, p-value < 2.2e-16
alternative hypothesis: true autocorrelation is greater than 0
DW = 0.23248>0.05，残差序列不存在自相关

残差的单整性检验

> m<-ur.df(residuals(s), type = c("none"), lags = 1，   
>  +                          selectlags = c("AIC"))
> summary(m)

############################################### 
# Augmented Dickey-Fuller Test Unit Root Test # 
############################################### 

Test regression none 


Call:
lm(formula = z.diff ~ z.lag.1 - 1 + z.diff.lag)

Residuals:
    Min      1Q  Median      3Q     Max 
-5.7814 -1.3393  0.1946  1.4806  5.1945 

Coefficients:
           Estimate Std. Error t value Pr(>|t|)  
z.lag.1    -0.13787    0.06857  -2.011   0.0499 *
z.diff.lag  0.22519    0.13969   1.612   0.1134  
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 2.22 on 49 degrees of freedom
Multiple R-squared:  0.097,	Adjusted R-squared:  0.06015 
F-statistic: 2.632 on 2 and 49 DF,  p-value: 0.08209


Value of test-statistic is: -2.0106 

Critical values for test statistics: 
     1pct  5pct 10pct
tau1 -2.6 -1.95 -1.61

-2.0106<-1.95,<-1.61，在拒绝域里，不存在单位跟，残差序列平稳

故存在协整关系
代表了一种长期均衡关系

残差图没有明显的时间趋势

误差修正模型

> diffk<<-diff(k_likelihood[11:63])
> diffg<-diff(g)
> error<-residuals(s)
> error.lag<-error[-(52:52)]
> ecm.reg1<-lm(diffk~diffg+error.lag)
> summary(ecm.reg1)

Call:
lm(formula = diffk ~ diffg + error.lag)

Residuals:
    Min      1Q  Median      3Q     Max 
-2.8531 -0.8762 -0.0453  1.0218  3.6047 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept) -1.74864    0.29075  -6.014 2.22e-07 ***
diffg       21.68558   10.16602   2.133   0.0379 *  
error.lag   -0.06900    0.04451  -1.550   0.1275    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 1.497 on 49 degrees of freedom
Multiple R-squared:  0.1347,	Adjusted R-squared:  0.09941 
F-statistic: 3.815 on 2 and 49 DF,  p-value: 0.02885
> dwtest(ecm.reg1)

	Durbin-Watson test

data:  ecm.reg1
DW = 2.2397, p-value = 0.7558
alternative hypothesis: true autocorrelation is greater than 0
残差序列也没有自相关

∆k_t=-1.74864+21.68558∆g_t-0.06900〖ecm〗_t
存在短期均衡

格兰杰因果关系检验==