Scribd
Upload
72 views

Tutorial 2 - Sol

This document provides a quantitative finance tutorial on ARMA processes and stationarity. It includes: 1) Discussion of key concepts like weak and strong stationarity as they relate to ARMA processes. 2) Exercises analyzing ARMA models to determine stationarity and invertibility, calculate means and variances, and derive autocovariance and autocorrelation functions. 3) Additional exercises showing properties of MA(1) processes and how combining independent processes can result in an MA(1) process.

Uploaded by

Vidaup40
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
72 views

Tutorial 2 - Sol

This document provides a quantitative finance tutorial on ARMA processes and stationarity. It includes: 1) Discussion of key concepts like weak and strong stationarity as they relate to ARMA processes. 2) Exercises analyzing ARMA models to determine stationarity and invertibility, calculate means and variances, and derive autocovariance and autocorrelation functions. 3) Additional exercises showing properties of MA(1) processes and how combining independent processes can result in an MA(1) process.

Uploaded by

Vidaup40
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Quantitative Finance

Tutorial 2
September 5, 2020

1 ARMA Processes and stationarity


1.1 Comment
1. The difference between weak and strong stationarity lies in the moments involved in the
independence analysis of a random variable distribution as function of time.
TRUE. Weak stationarity is based around the first two central moments of the random
variable. These should not vary across time for the variable to be categorized as stationary.
In a strong stationary process, all moments must be time invariant.
2. Every M A(1) process can be represented as a AR(∞) process.
FALSE. This holds true only if the process is invertible and thus the roots of the M A(1)
series are less than 1 in absolute value.
3. A finite M A process is always stationary.
TRUE. Given a white noise error term.

Yt = εt − βt εt−1 − · · · − βq εt−q
q
!
X
V(Yt ) = 1 + βi σ 2
2

i=1

E(Yt ) = 0

1.2 Exercises
1. An insurance company uses the following model to forecast accidents.

Yt = 0.3Yt−1 − 0.7µt−1 + µt

Where Yt is the number of accidents and µt ∼ N (0, 1)

(a) Identify the model. Is it stationary? Is it invertible? Calculate its mean and variance.
ARMA(1,1)
Stationary: |0.3| < 1
Invertible: | − 0.7| < 1
E(Yt ) = 0
V(Yt ) = (0.3)2 V(Yt ) + σε2 + (0.7)2 σε2 − 2(0.3)(0.7)σε2
V(Yt ) = 1.175
(b) Calculate the Autocovariance and the Autocorrelation functions from 0 to 3.
γ0 = 1.175
γ1 = C(Yt , Yt−1 )
γ1 = 0.3γ0 − 0.7σε2 = −0.347
γ2 = 0.3γ1 = −0.104
γ3 = 0.3γ2 = −0.031
γk = 0.3γk−1 ; ∀k > 1

1
ρ0 = γγ00 = 1
ρ1 = γγ10 = −0.295
ρ2 = γγ20 = −0.3 γγ10 = −0.3ρ1 = −0.038
ρ3 = −0.3ρ2 = −0.026
ρk = −0.3ρk−1

P∞the model as a AR(∞).


(c) Rewrite
Yt = i=1 0.7i−1 (0.3 − 0.7)Yt−i + µt

P∞the model as a M A(∞).


(d) Rewrite
Yt = i=1 0.3i−1 (0.3 − 0.7)µt−i + µt
2. Consider the following M A(1) process: yt = εt + θεt−1 . Show that the first autocorrelation
satisfies |ρ| ≤ 0.5.
θ
ρ1 = (1+θ) 2

Given that the process is stationary |θ| < 1


Testing the limits
ρ1 = 0.5; θ = 1
ρ1 = −0.5; θ = −1

3. Consider the following MA proccesses: y1,t = (1 + θ1 L)ε1,t and y2,t = (1 + θ2 L)ε2,t . Where
ε1,t and ε2,t are white noise independendent shocks. Show that yt = y1,t + y2,t follows a
M A(1) process.
γ0(i) = (1 + θi2 σi2 )
γ1(i) = θi σi2
γk(i) = 0; k > 1
Since the two proccesses are independent then
γ0 = (1 + θ12 σ12 ) + (1 + θ22 σ22 )
γ1 = θ1 σ12 + θ2 σ22
γk = 0; k > 1
Thus, follows a regular M A(1) process.
4. Consider the following ARM A(2, 1): yt = φ1 yt−1 + φ2 yt−1 − φ1 φ2 yt−2 + εt − φ2 εt−1 . Where
|φ1 | < 1 and |φ2 | < 1. Find the Autocovariance and Autocorrelation functions of yt .
yt − φ1 yt−1 − φ2 yt−1 + φ1 φ2 yt−2 = εt − φεt−1
(1 − (φ1 + φ2 )L + φ1 φ2 L2 )yt = (1 − φ2 L)εt
(1 − φ1 L)(1 − φ2 L)yt = (1 − φ2 L)εt
yt = φ1 yt−1 + εt
γ1 = φ1 γ0
γ2 = φ21 γ0
γk = φk1 γ0
ρk = φk1 ; ∀k >= 1

You might also like