Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Time Series Project Group F 

 
Question 1: Simulation of GARCH(2,3)  
 
Part A: Boundary Conditions α1 = 0 ; α2 = 0 ; β 1 = 0 ; β 2 = 0 ; β 3 = 0  
 
Time Series Plot 
 

 
 
Part B: Boundary Conditions α1 = 0.1 ; α2 = 0 ; β 1 = 0.1 ; β 2 = 0 ; β 3 = 0  
 
Time Series Plot 
Part C: Boundary Conditions α1 = 0.2 ; α2 = 0.2 ; β 1 = 0.1 ; β 2 = 0.1 ; β 3 = 0  
 
Time Series Plot 

Part D: Boundary Conditions α1 = 0.2 ; α2 = 0.2 ; β 1 = 0.2 ; β 2 = 0.19 ; β 3 = 0.19  


 
Time Series Plot 
 
Question 2: Plot of ACF for every boundary condition 
 
​Part A: Boundary Conditions α1 = 0 ; α2 = 0 ; β 1 = 0 ; β 2 = 0 ; β 3 = 0  

ACF of xt   ACF of ||xt ||   ACF of x2t   


 
Part B: Boundary Conditions​ α1 = 0.1 ; α2 = 0 ; β 1 = 0.1 ; β 2 = 0 ; β 3 = 0  

ACF of xt   ACF of ||xt ||   ACF of x2t   


 
Part C: Boundary Conditions α1 = 0.2 ; α2 = 0.2 ; β 1 = 0.1 ; β 2 = 0.1 ; β 3 = 0  

ACF of xt   ACF of ||xt ||   ACF of x2t   


Part D: Boundary Conditions​ α1 = 0.2 ; α2 = 0.2 ; β 1 = 0.2 ; β 2 = 0.19 ; β 3 = 0.19  
ACF of xt   ACF of ||xt ||   ACF of x2t   
 
Question 3: Analyzing Macroeconomic Data with VAR Model 
 
Part 1 ​:​ We took the GDP and Per Capita Consumption of USA and did some data 
cleaning and removed missing values 
 
Part 2:​Growth Rate of Consumption Per Capita  
 
1st Iteration 

 
 
ADF Stationarity Test : Null Hypothesis is Rejected as p value is high and hence it is 
non stationary 
 

 
 
Taking Lag Difference of growth Rate after 2 iterations 
 
Final Growth Rate of Consumption Per Capita 
 

 
 
ADF Stationarity Test: Final plot is Stationary 
 
Growth Rate of GDP Per Capita - 1​ st iteration  
 

 
ADF Test for Stationarity:  
 

 
Null Hypothesis is Rejected 
 
Taking Lag Difference of growth Rate after 1 iteration 
Final Growth Rate plot 
ADF Stationarity Test: Final plot is Stationary 
 

 
 
 
Part 3: ACF & PACF of Consumption per Capita 

 
ACF & PACF of GDP per Capita 

 
 
 
 
 
 
 
 
Part 4: Estimate ARMA(p, q) processes separately for each of the two series: 
 
Estimating ARMA for Consumption Per Capita 
 
Table of AIC & BIC with respective orders 

   
 
Here we see that at p=1 & 1=1 the value of AIC is lowest therefore this model is 
chosen 
Estimating ARMA for GDP Per Capita 
 
Table of AIC & BIC with respective orders 

 
We conclude that at p=1,q=0 OR p=2, q=1 our model is giving lowest AIC 
therefore either of the model can be selected for it.  
Part 5:Fit VAR(p) for p = 0, . . . , 4. Compute and store AIC and BIC.  
 
We fitted the VAR models and find out model 2 (var_m2) is better with the help of 
optimal lag selection method and checked stability of model as roots of model 2 
(var_m2) lying inside the unit circle​.

Then we performed 4 diagnostic test to check if our model selection is right or 
not:   

1. Serial Correlation- P
​ assed this test 
2. Heteroscedasticity - ​Passed this test 
3. Normal Distribution of the Residuals- ​Passed as Residuals are normally 
distributed 
4. Testing for structural breaks in the Residuals- there are no points in these 
graphs (attached image) that exceeds the two red lines- ​Passed  

 
 

 
 
Part 6: Does consumption Granger cause gdp or the other way round?  
 
Performing the Granger Causality Test on both GDP and Consumption Per Capita we 
obtain the following: 
 
● GDP Per Capita does not causes Consumption Per Capita as p value is less 
than 0.05 and hence cannot reject the null hypothesis 
● On the other hand, Consumption Per Capita does not causes GDP Per Capita 
as p value is less than 0.05 and hence cannot reject the null hypothesis 
●  
Therefore both variables do not cause Granger Causality amongst each other 
 
 
 
 
Proof and Methodology 
 

 
 

 
 
   
 
 
Conclusion​: We find that both GDP and Consumption Per Capita are almost in 
correlation with each other  
 
 

You might also like