North-West University
School of Mathematical and Statistical Sciences
STTN326: Analysis of Dependent Data
Tutorial 2
Study Unit 2: Models for Stationary Time Series
Instructions
• Show all calculations.
• State all theorems used.
• Simplify your answers where possible.
• This tutorial will be discussed during the tutorial session.
Question 1 (Theory)
Answer the following questions.
a) Define a linear process.
b) State Wold’s Decomposition Theorem.
c) Explain the difference between an AR(p) process and an MA(q) process.
d) What are the three properties of a white noise process?
e) Explain why stationarity is important before fitting an AR model.
Question 2 (Model Identification)
Identify each of the following models.
a)
Xt = 0.8Xt−1 − 0.3Xt−2 + Wt
b)
Xt = Wt + 0.4Wt−1 − 0.2Wt−2
c)
Xt = 0.5Xt−1 + Wt + 0.7Wt−1
d)
Xt = −0.6Xt−1 + 0.4Xt−3 + Wt
State the order of each process.
1
Question 3 (Backshift Operator)
Express each of the following using backshift notation.
a)
Xt = 0.6Xt−1 − 0.2Xt−2 + Wt
b)
Xt = Wt + 0.5Wt−1 + 0.3Wt−2
c) Write the first-order difference operator using the backshift operator.
d) Show that
(1 − B)2 Xt = Xt − 2Xt−1 + Xt−2 .
Question 4 (Stationarity)
Determine whether each AR process is stationary.
a)
(1 − 0.7B)Xt = Wt
b)
(1 − 1.2B)Xt = Wt
c)
1 1
Xt = Xt−1 + Xt−2 + Wt
2 4
d)
4
Xt = Xt−2 + Wt
9
For each process,
(a) write the AR operator;
(b) determine the characteristic equation;
(c) calculate the characteristic roots;
(d) conclude whether the process is stationary.
2
Question 5 (Inverse AR Operator)
Consider the AR(1) process
(1 − 0.5B)Xt = Wt .
(a) Find the inverse AR operator.
(b) Write the first five terms of the MA(∞) representation.
(c) Identify
ψ0 , ψ1 , ψ2 , ψ3 , ψ4 .
Question 6 (Invertibility)
Consider the MA(1) process
1
Xt = Wt + Wt−1 .
2
(a) Write the MA operator.
(b) Determine the characteristic equation.
(c) Determine the characteristic root.
(d) Is the process invertible?
(e) Obtain the first five terms of the AR(∞) representation.
Question 7 (Equivalent MA Models)
Consider
Xt = Wt + 3Wt−1 ,
and
1
Yt = Wt + Wt−1 .
3
(a) Compute the lag-one autocorrelation for both models.
(b) Are the two processes equivalent?
(c) Explain why invertibility is required for MA models.
3
Question 8 (ARMA Models)
Consider
(1 − 0.6B)Xt = (1 + 0.4B)Wt .
(a) Identify the model.
(b) Determine whether the model is stationary.
(c) Determine whether the model is invertible.
(d) State the conditions required for an ARMA model to be both stationary and
invertible.
Question 9 (Challenge Question)
Suppose
Xt = 1.5Xt−1 − 0.56Xt−2 + Wt .
(a) Determine the AR operator.
(b) Obtain the characteristic polynomial.
(c) Determine the characteristic roots.
(d) Is the process stationary?
(e) If stationary, explain how the process can be written as a linear process.
Question 10: Simple (Non-Redundant) Forms of
ARMA Models
Find the simple (non-redundant) form of each of the following ARMA models.
Clearly show all your working.
1.
3 3
Xt = Xt−1 + Wt + Wt−1 .
4 4
2.
1 1
Xt = − Xt−1 + Wt − Wt−1 .
3 3
3.
2 1 1 2
Xt = Xt−1 − Xt−2 + Wt + Wt−1 − Wt−2 .
5 5 5 5
4.
1 1 1 1
Xt = − Xt−1 + Xt−2 + Wt − Wt−1 + Wt−2 .
2 4 2 4