Scribd
Upload
130 views

Week6 Assignment Solutions

The document contains 10 multiple choice questions related to time-series analysis concepts such as stationarity, autocorrelation, Fourier analysis, and filtering. For each question, the correct answer is provided along with a brief explanation of the reasoning. Key concepts covered include identifying appropriate ARIMA models, transformations to achieve stationarity, frequency components in Fourier representations, properties of autocorrelation functions, and computing cross-covariances from known autocovariances.

Uploaded by

vicky.sajnani
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
130 views

Week6 Assignment Solutions

The document contains 10 multiple choice questions related to time-series analysis concepts such as stationarity, autocorrelation, Fourier analysis, and filtering. For each question, the correct answer is provided along with a brief explanation of the reasoning. Key concepts covered include identifying appropriate ARIMA models, transformations to achieve stationarity, frequency components in Fourier representations, properties of autocorrelation functions, and computing cross-covariances from known autocovariances.

Uploaded by

vicky.sajnani
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 14

Applied Time-Series Analysis

Arun K. Tangirala

Week 6 Assignment Solutions


1. Consider the model object given in w6 q1.Rdata, fit to an arbitrary process. Is this model
admissible?
a. Yes, no other model would be possible.
b. Yes, but another possible model would be ARMA(1, 1)
c. No, and an appropriate model would be ARIMA(0, 1, 0)
d. No, and an appropriate model would be ARMA(0, 1)
Answer: (c)

load("w6_q1.Rdata")
>print(model)

Call:
arima(x = x, order = c(1, 0, 0))
Coefficients:
ar1 intercept
0.9974 -8.0649
s.e. 0.0020 9.0732

sigma^2 estimated as 0.9387: log likelihood = -1391.32, aic = 2786.65

We see that the confidence interval estimated for the model coefficient crosses 1. Thus the
given model is non-stationary and as a consequence, inadmissible. Extending this reasoning it
is obvious that the given model is an AR model fit to a pure integrating process. Hence, an
ARIMA(0,1,0) is the most appropriate model.
2. For the series x[k] given in w6 q2.Rdata, what is an appropriate transformation g(.) that
makes the data stationary?
p
a. g(x[k]) = x[k]
b. g(x[k]) = log(x[k])
1
c. g(x[k]) = p
p x[k]
d. g(x[k]) = x[k]
Answer: (b)

The appropriate Box-Cox transformation for the data is the log transformation as this renders
the data stationary.
3. For the transformed data in Question 2, the most appropriate model is .
a. ARMA(1,0) model
b. ARMA(1,1) model
c. ARMA(3,0) model
d. ARMA(0,2) model
Answer: (c)

The most appropriate model for the transformed data is ARMA(3,0) model.
4. For the following continuous-time signal



 1, 0 ≤ t ≤ 1/2

x(t) = −1, 1/2 ≤ t ≤ 1


0,

otherwise

which of the following statements is / are TRUE?


a. It is an energy signal with Exx = 1
b. It is a power signal with Pxx = 1
c. It is a power signal with Pxx = 0.5
d. It is an energy signal with Exx = 0.5

Answer: (a)
Since the signal is of finite duration, it has finite energy and hence is an energy signal. Energy
signals by definition have zero power. The energy can be calculated as
Z ∞
Exx = |x(t)|2 dt
−∞
Z 1
= |x(t)|2 dt
0
=1
5. What are the valid frequency components present in the Fourier-series representation of
the given periodic signal:

x[k] = sin(2πf0 k) + cos2 (2πf0 k)

where f0 = 0.2?
a. 0.2, 0.4
b. 0.4, 0.6
c. 0.8, 0.4
d. 0.2, 1.2
Answer: (a)

The given signal can be re-written as


(1 + cos(2π(2f0 )k))
x[k] = sin(2πf0 k) +
2
We observe that the frequency components present are f0 and 2f0 , i.e., 0.2 and 0.4.
6. For the given continuous time signal
x(t) = cos(t) + sin(2t),
the value of average power Pxx is (use Parseval’s relation) .
a. 2
b. 0
c. 0.5
d. 1

Answer: (d)
The given signal can be written as
eit + e−it e2it − e−2it
x(t) = +
2 2i
1 it 1 −it 1 2it 1 −2it
= e + e − ie − ie
2 2 2 2
12 12 12 12
Pxx =| | +| | +| | +| | =1
2 2 2 2
πn 3πn
 
7. The Fourier coefficients of the periodic signal x[k] are given by cn = cos 4 + sin 4 .
The signal x[k] with period Np = 8 is .
a. {0, 4, 0, 4i, 0, −4i, 0, 4}
b. {0, −4, 4i, 0, 0, −4i, −4}
c. {0, 4, −4i, 0, 0, −4i, 4}
d. {0, −4, 4i, 0, 0, −4i, −4}

Answer: (a)
Given that the fundamental period Np = 8 samples. Therefore series x[k] is
7
X
x[k] = Cn ej2πnk
n=0

The periodic signal is

x[k] = {0, 4, 0, 4i, 0, −4i, 0, 4} starting from k = 0


8. Which of the following ACF signatures represents that of a deterministic, periodic signal?

(a) (b)

(c) (d)
Answer: (a)

For a deterministic, periodic signal the ACVF is also a periodic at all lags. The only option that
reflects this property is (a).
9. Two stationary signals v[k] and w[k] are known to be correlated. Then, the following
always holds:
a. ρvw [0] 6= 0
b. ρ̂vw [l] > 0, ∀l
2
c. σvw [l] ≤ σvv [l]σww [l], ∀l
d. ρ̂[1] = 1

Answer: (c)
10. The auto-covariance function of an input x[k] to a random process with known transfer
function G(q) = 1 + 0.4q −1 is known to be σxx [0] = 8, σxx [1] = 2, σxx [l] = 0, ∀l > 1.
The cross-covariance between the input and output at lag l = 0 is .
a. 4.4
b. 8.8
c. 5.2
d. 10.88

Answer: (b)
The given transfer function is
y[k]
G(q) = = 1 + 0.4q −1
x[k]
y[k] = x[k] + 0.4x[k − 1]
σxy [0] = σxx [0] + 0.4σxx [1]
= 8.8

You might also like