D 57 MTW 11 Solwb

UNIVERSITY OF TORONTO SCARBOROUH
STAD57H3 Time Series Analysis

Midterm Test
March, 2011
Duration: one hour and fifty minutes
Please do NOT open this booklet until you are asked to do so.
There are 9 pages including this page and t tables. Please check to see if you have all the pages.
Aids allowed: You are allowed to use the textbook (Time Series Analysis and Its Applications
with R Examples by R. H. Shumway and D. S. Stoffer), the class notes and a non-
communicating calculator. No other material will be allowed during the test.
All your work must be presented clearly in order to get credit. Just an answer with no other work
shown will only qualify for zero credit. Show your work and answer in the space provided, in
ink. Pencil may be used, but then any re-grading will NOT be allowed.
You will get a mark out of 60 for this test.
Last name:____________________
First name:____________________
Other names:___________________
Student number: _________________
Page 1 of 9
1) [ 6 points] Let { X t } be a weakly stationary time series and let’s assume that we have made
three observations from this series i,e, n = 3. Let = 5, = 3 and = 4. Use this data to
estimate γ (1). (i.e. calculate γˆX (1) )
Sol
1 n−h
γˆX (h) = ∑ ( xt −x )( xt + h − x )
n t =1
1 n −1
γˆX (1) = ∑ ( xt −x )( xt + h − x )
n t =1
1 2
= ∑ ( xt −x )( xt + h − x )
3 t =1
1
= ( ( 5 − 4 ) (3 − 4) + (3 − 4)((4 − 4) )
3
1
=−
3
Page 2 of 9
2) Consider the AR(2) process given by X t = X t −1 − 0.5 X t − 2 + wt .
a) [5 points] Is this process weakly stationary? Give reasons for your answer.
Sol
The characteristic equation is
The roots of this equation (regarding B as a variable) are complex, namely, 1± i. As the rnodulus
of both roots exceeds one, the roots are both outside the unit circle and so the process is
stationary.
b)[ 10 points] Calculate Corr ( X 1 , X 2 ) and Corr ( X 5 , X 7 ) .
Sol:
Since this process is stationary Corr ( X 1 , X 2 ) = ρ (1) and Corr ( X 5 , X 7 ) = ρ (2)
p
Yule-Walker equations γ (h) − ∑ φiγ (h − i ) = σ w2 I (h = 0) .
i =1
h=1
2
γ (1) − ∑ φiγ (1 − i ) = 0
i =1
i.e. γ (1) − γ (0) + 0.5γ (1) = 0

1.5γ (1) − γ (0) = 0
γ (1) 1 2
ρ (1) = = =
γ (0) 1.5 3
h= 2
2
γ (2) − ∑ φiγ (2 − i ) = 0
i =1
i.e. γ (2) − γ (1) + 0.5γ (0) = 0

ρ (2) − ρ (1) + 0.5 ρ (0) = 0
2 1
ρ (2) = ρ (1) − 0.5ρ (0) = − 0.5 =
3 6
Page 3 of 9
3) [7 points] Consider the stationary ARMA(1, 1) process given by
X t = 0.5 X t −1 + wt − 0.3wt −1 .
Since the process is (weakly) stationary, it has the Wold representation. Calculate the first four
coefficients of the Wold representation (i.e. ψ 0 , ψ 1 , ψ 2 , and ψ 4 , )
Sol
ψ ( B ) = (1 − 0.3B )(1 − 0.5 B ) −1

= (1 − 0.3B )(1 + 0.5 B + 0.52 B 2 + 0.53 B 3 + )
= 1 + 0.2 B + 0.1B 2 + 0.05 B 3 +
Page 4 of 9
4) [ 6 points] A stationary process { X t } has ρ (1) = 0.97 and ρ (2) = 0.90. Calculate φ2,2 (i.e. the
partial autocorrelation at lag 2).
Sol
−1
 1 0.97   0.97   1.64129 
φ2,2 is the second element of the vector    = 
 0.97 1   0.90   -0.69205 
MTB > Read 2 2 m1.

DATA> 1 0.97
DATA> 0.97 1
2 rows read.
MTB > Invert M1 m2.
MTB > print m2
Data Display
Matrix M2
16.9205 -16.4129
-16.4129 16.9205
.
MTB > Read 2 1 m3.
DATA> 0.97
DATA> 0.90
2 rows read.
MTB > print m3
Data Display
Matrix M3
0.97
0.90
MTB > Multiply M2 M3 m5.

MTB > print m5
Data Display
Matrix M5
1.64129
-0.69205
MTB >
Or can use the Yule-Walker equations
∑α
i =1
*
ih γ ( j − i ) = γ ( j ), j = 1,… , h can use ρ’s in place of γ’s.
Page 5 of 9
In this question h = 2 and so
j =1
α12* ρ (0) + α 22
*
ρ (1) = ρ (1)
α12* + 0.97α 22
*
= 0.97
j= 2
α12* ρ (1) + α 22
*
ρ (0) = ρ (2)
0.97α12* + α 22
*
= 0.97
*
and φ2,2 = α 22
Page 6 of 9
5) [ 6 points] A time series { X t } is given by X t = 2 wt + wt −1 + vt where wt and vt are
uncorrelated (i.e. Cov( ws , vt ) = 0 for all s, t.) white noise processes, each with mean 0 and
variance 3.
Calculate the cross-correlation function between {wt } and { X t } .
Sol
V ( X t ) = 4V ( wt ) + V ( wt −1 ) + V ( vt ) = 4 × 3 + 3 + 3 = 18
γ w, x (h) = Cov( wt , X t + h )
= Cov( wt , 2 wt + h + wt + h−1 + vt + h )
= 2Cov( wt , wt + h ) + Cov( wt , wt + h−1 )
 6 if h = 0

=  3 if h =1
0 otherwise

Corr ( wt , X t + h )
ρ w , x ( h) =
VwVX t t+h
6 / 3 ×18 if h = 0

=  3 / 3 ×18 if h =1
0 otherwise

Page 7 of 9
6) [ 6 points] A time series { X t } is given by X t = U cos(ct ) + V sin(ct ) where U and V are iid
N(0, σ ), σ > 0 and c > 0 is a constant. Calculate Cov( , ) for h = 0, ±1, ±2, ⋯. Is { X t }
a weakly stationary process?
( Hint: cos( A − B ) = cos( A) cos( B ) + sin( A) sin( B ) )
Sol
Page 8 of 9
7) Consider the time series model given by X t = 0.5 + 0.75 X t −1 + wt where wt are i.i.d.
(0, σ ). You may assume that this process is weakly stationary. [Note: This is not exactly the
AR(1) process we discussed in class, but the necessary formulas can be developed in a similar
manner.]

a) [4 points] If =1.5, find a one-step ahead forecast (i.e. )
Sol
( )
X tt+l = E X t + l | X t
= E ( 0.5 + 0.75 X t + l −1 + wt + l | X t )
 0.5 + 0.75 X t if l = 1
= t
0.5 + 0.75 X t +l −1 if l ≥ 2

And so = 0.5+0.75*1.5 = 1.625

b) [ 6 points] If =1.5, find a 10-step ahead forecast (i.e. )
Sol
X tt+ 2 = 0.5 + 0.75 X tt+1
= 0.5 + 0.75 ( 0.5 + 0.75 X t )
= 0.5 + 0.5 × 0.75 + 0.752 X t
X tt+3 = 0.5 + 0.75 X tt+ 2

(
= 0.5 + 0.75 0.5 + 0.5 × 0.75 + 0.752 X t )
= 0.5 + 0.5 × 0.75 + 0.5 × 0.752 + 0.753 X t
( )
X tt+l = 0.5 + 0.5 × 0.75 + 0.5 × 0.752 + + 0.5 × 0.75l −1 + 0.75l X t
0.5(1 − 0.75l )
= + 0.75l X t
1 − 0.75
= 2(1 − 0.75l ) + 0.75l X t
X tt+10 = 2(1 − 0.7510 ) + 0.7510 × 1.5 = 1.971843243

c) [ 4 points] Calculate E ( X t ) for this process.
Sol
E ( X t ) = µ ∀t since the process is stationaty .
X t = 0.5 + 0.75 X t −1 + wt ⇒ EX t = 0.5 + 0.75 EX t −1 + Ewt

=µ =µ =0
µ = 0.5/(1-0.75) = 2
Page 9 of 9

D 57 MTW 11 Solwb

Uploaded by

Copyright:

Available Formats

D 57 MTW 11 Solwb

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

D 57 MTW 11 Solwb

Uploaded by

Copyright:

Available Formats

UNIVERSITY OF TORONTO SCARBOROUH

STAD57H3 Time Series Analysis

You will get a mark out of 60 for this test.

Student number: _________________

b)[ 10 points] Calculate Corr ( X 1 , X 2 ) and Corr ( X 5 , X 7 ) .

i.e. γ (1) − γ (0) + 0.5γ (1) = 0

i.e. γ (2) − γ (1) + 0.5γ (0) = 0

ψ ( B ) = (1 − 0.3B )(1 − 0.5 B ) −1

MTB > Read 2 2 m1.

MTB > print m3

MTB > Multiply M2 M3 m5.

Or can use the Yule-Walker equations

X tt+3 = 0.5 + 0.75 X tt+ 2

X tt+10 = 2(1 − 0.7510 ) + 0.7510 × 1.5 = 1.971843243

X t = 0.5 + 0.75 X t −1 + wt ⇒ EX t = 0.5 + 0.75 EX t −1 + Ewt

You might also like