Scribd
Upload
7 views

Assignment 1

Uploaded by

liuzhe20030124
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
7 views

Assignment 1

Uploaded by

liuzhe20030124
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Time Series Analysis

ASSIGNMENT ONE

The due date for this assignment is Tuesday 22 October.


This assignment accounts for 10% of the total course mark. (Remarks for each problem is
indicated at the end of the question. This will be translated to the final course grade
proportionally.)

Some problems need to be completed by R. Please include your code at the end of the
assignment as Appendix.

Be sure to include your name and ID number.

Please hand in on time. Otherwise there will be a 50% discount on your remark.

1
1. For the MA(2) model
𝑟! = 𝑎! + 𝜃" 𝑎!#" + 𝜃$ 𝑎!#$ , where 𝑎! is white noise (0, 𝜎 $ )
(a) Is 𝑟! covariance-stationary? Compute its mean and variance (5’)
(b) Prove its ACF cuts off at lag 2. (5’)
(c) Compute the 𝑙-step ahead forecast, the forecast errors and their variances.
What conclusion do you get? (5’)
(d) Explore whether the PACF of 𝑟! has cut off pattern or not. (5’)

2. Consider the AR(2) model


𝑟! = 𝜙" 𝑟!#" + 𝜙$ 𝑟!#$ + 𝑎! , where 𝑎! is white noise (0, 𝜎 $ ).
(a) What is the stationarity condition in terms of 𝜙" , 𝜙$ ? Draw the
stationarity region in the 2-dimentional plane of 𝜙" , 𝜙$ . (5’)
(b) Use the conclusion in (a) to decide whether the following AR(2) model
is covariance stationary or not.
𝑟! = 1.1𝑟!#" − 0.18𝑟!#$ + 𝑎! , where 𝑎! is white noise (0,1).
If so, compute its auto-covariance. (5’)
(c) Suppose 𝜌" = 0.5 and 𝜌$ = 0.3. Solve 𝜙" , 𝜙$ . (5)

3. Consider the following two MA(2) models,


(1) 𝑍"! = (1 + 2𝐵)(1 + 0.2𝐵)𝑎! , 𝑤ℎ𝑒𝑟𝑒 𝑎! ~𝑖. 𝑖. 𝑑. 𝑁 (0,2).
(2) 𝑍$! = (1 + 0.5𝐵)(1 + 0.2𝐵)𝑏! , 𝑤ℎ𝑒𝑟𝑒 𝑎! ~𝑖. 𝑖. 𝑑. 𝑁(0,8).
(a) Are the models invertible? (5’)
(b) Compute the variance, auto-covariance and ACF for both models. What
conclusion do you get? (5’)

4. Consider

$
𝑔% = A𝛼 + 𝛽𝑔%#" ∙ 𝜀% , where 𝜀% ~𝑖𝑖𝑑(0,1), 𝛼 > 0, 0 < 𝛽 < 1,
(a) Is 𝑔% an independent series? Is 𝑔% serially correlated? Is 𝑔% covariance-
stationary? (7’)

2
$
(b) Suppose now you are at time t, how would you forecast 𝑔!&" and 𝑔!&" ?
What conclusion do you get? (8’)

5. For an ARMA(1, 2) model


(1 − ∅𝐵 )𝑌! = (1 − 𝜃" 𝐵 − 𝜃$ 𝐵 $ )𝑎! , where 𝑎! ~𝑖. 𝑖. 𝑑. 𝑁 (0, 𝜎 $ ),
(a) When can this model be reduced to an MA(1) model? (5’)

Assuming ∅ = 0.7, 𝜃" = 0.2, 𝜃$ = −0.35, 𝜎 $ = 2, simulate 200 observations


for 𝑌! .

(b) Tentatively specify an ARMA model for the simulated data. How
would you estimate the model? (5’)

(c) Produce forecast for the series for 10 periods based on the model you
selected in (b). Plot the forecasts and the forecast error. (5’)

6. Suppose {𝑋! } and {𝑌! } are both stationary and they are independent with
each other. They share the same mean and auto-covariance function 𝛾' .
Define
𝑋 , 𝑖𝑓 𝑡 = 2𝑛 + 1,
𝑍! = O !
𝑌! , 𝑖𝑓 𝑡 = 2𝑛.
Is {𝑍! } stationary? Prove your conclusion. (5’)

7. Suppose {𝜀! } is WN(0, 𝜎 $ ), 𝑋( = 0. For 𝑋! = 𝑋!#" + 𝜀! , compute the


correlation coefficient 𝜌(𝑡, 𝑠) between 𝑋! and 𝑋) . (5’)

8. Consider the time series: x-series.xlsx

(a) Plot the series and comment on stationarity and general behaviour of the
series. (5’)
(b) Tentatively specify (or identify) a time series model for the series.
Briefly explain why. (5’)
3
(c) Estimate the model you specified in question (b) and do some diagnostic
check. Comment on the adequacy of the model. (5’)

You might also like