University College Dublin

An Coláiste Ollscoile, Baile Átha Cliath

Autumn Trimester Exam, 2022

STAT30010 / STAT40700

TIME SERIES

Professor R. Killick
Associate Professor Adam Butt, FIAA
Mr. Niall Franklin, BSc (Maths)
Associate Professor Edward Cox
Dr. Wagner Barreto-Souza∗
Time Allowed: 2 hours

Instructions for Candidates

Full marks will be awarded for complete and correct answers to all six questions.
For full 100/100 marks, you must show clearly all steps and computations.

Notes for Invigilators

Non-programmable calculators may be used during this examination.
No notes or books of tables or formulae are to be issued to candidates.

© UCD 2022/2023 1 of 5
Notation: Throughout this paper, WN(0, σ 2 ) denotes a white noise with mean 0 and
variance σ 2 .

1. Let {Wt }t∈Z ∼ WN(0, σ 2 ) and define the time series {Xt }t∈Z by

Xt = Wt + θWt−1 , t ∈ Z,

where θ is a real constant.

(a) Compute the mean function of {Xt }t∈Z . [5]

(b) Compute the variance function of {Xt }t∈Z . [5]
(c) Compute the autocovariance function of {Xt }t∈Z . [5]
(d) For |θ| < 1, show that {Wt }t∈Z can be represented in terms of Xt , Xt−1 , . . ..
X∞
zj .
1
Hint: For a complex number z with |z| < 1, we have that = [5]
1−z j=0
[Total 20 marks]

2. Assume that {Xt }t∈Z is a time series defined by

Xt = θ0 + θ1 t + Wt , t ∈ Z,

where {Wt }t∈Z ∼ WN(0, σ 2 ), and θ0 and θ1 are real constants.

(a) Argue that {Xt }t∈Z is not stationary. [5]

(b) Propose a transformation of the time series {Xt }t∈Z to achieve stationarity.
[5]
[Total 10 marks]

3. Consider an ARMA process, say {Xt }t∈Z , given by

4 9
Xt = Xt−1 + Xt−2 + Wt + 2Wt−1 , t ∈ Z,
10 20
where {Wt }t∈Z ∼ WN(0, σ 2 ).

(a) Identify the autoregressive and moving average polynomials associated with
{Xt }t∈Z . [5]
(b) Is {Xt }t∈Z stationary? Justify your answer. [5]
(c) Is {Xt }t∈Z causal? Justify your answer. [5]
(d) Is {Xt }t∈Z invertible? Justify your answer. [5]

[Total 20 marks]

© UCD 2022/2023 2 of 5
 4. Let {Xt }t∈Z be a causal autoregressive process given by Xt = φXt−2 + Wt , with
{Wt }t∈Z ∼ WN(0, σ 2 ). Assume that the following sample moments were obtained
after observing X1 , . . . , X200 : γ
b(0) = 6 and ρ
b(2) = 0.5 (remember the notation
γ(0) = Var(Xt ) and ρ(2) = corr(Xt , Xt−2 )). By using the causal assumption, find
the method of moments estimates for the parameters φ and σ 2 .

Hint: You can use the fact that E(Xt−j Wt ) = 0 due to the causality of {Xt }t∈Z ,
for j ≥ 1.
[Total 10 marks]

5. A time series x1 , . . . , x200 is observed. Figure 1 shows the plot of a such a time
series (to the left) with its respective sample ACF (middle) and PACF (to the
right) plots.

(a) Does this time series appear stationary? Propose a model for this time
series. Justify your answer. [10]
Series x Series x
1.0
4

0.6
0.8
2

0.4
0.6

Partial ACF
ACF
x
0

0.4

0.2
0.2
−2

0.0
0.0
−4

0 50 100 150 200 0 5 10 15 20 5 10 15 20

Time Lag Lag

Figure 1: Plot of x1 , . . . , x200 (to the left) with its respective sample ACF (middle) and
PACF (to the right) plots

© UCD 2022/2023 3 of 5
 (b) A model has been fitted to these data and the R output is provided below.
Write down the fitted equation model and discuss about the significance
of lags included. What is the estimate of the variance of the white noise?
[10]
R output:
##
##Call:
##arima(x = x, order = c(1, 0, 1))
##
##Coefficients:
## ar1 ma1 intercept
## 0.3602 0.4101 0.4635
##s.e. 0.0901 0.0817 0.1581
##
##sigma^2 estimated as 1.038: log likelihood = -287.82, aic = 583.64

(c) The diagnostic plots associated with the fitted model in letter (b) are provided
in Figure 2. Is the model well fitted? Comment on all the plots. [10]
Standardized Residuals
2
1
0
−3 −2 −1

0 50 100 150 200

ACF ofTime
Residuals
0.8
ACF
0.4 0.0

0 5 10 15 20
Lag
p values for Ljung−Box statistic
0.0 0.2 0.4 0.6 0.8 1.0
p value

2 4 6 8 10
lag

Figure 2: Diagnostics for the fitted model to the time series x1 , x2 , . . . , x200 .

[Total 30 marks]

© UCD 2022/2023 4 of 5
 6. Consider a VAR(1) process {(X1,t , X2,t )} defined by

X1,t = 2X1,t−1 + X2,t−1 + W1,t ,

X2,t = X1,t−1 + 2X2,t−1 + W2,t ,

where {(W1,t , W2,t )} is a bivariate white noise with mean vector 0 and covariance
matrix Σ. Is {(X1,t , X2,t )} a stationary process? Justify your answer.
[Total 10 marks]

—o0o—

© UCD 2022/2023 5 of 5