Tutorial Chapter 1 and 2 Memo-1
Tutorial Chapter 1 and 2 Memo-1
Tutorial Chapter 1 and 2 Memo-1
Chapter 1 and 2
Question 1 a) For a stochastic process { X t : t ∈ [0, ∞)} define the Markov property.
b) What is a stochastic process having the Markov property with a discrete parameter space
called? Markov chain
c) What is a stochastic process having the Markov property with a continuous parameter space
called? Markov Jump process
Question 2 Consider a branching process with X 0 = 1 i.e. the zero-th generation has only
one element. Let µ be the expected number of offspring of any element and let σ 2 be the
variance of the number of offspring of any element. Let X n be the size of the n th generation.
a) Show that E[ X n ] = µ n .
c) What can you say about the expected size of the n th generation as n → ∞ if µ < 1 ?
Question 3 What does it mean if a stochastic process { X t : t ∈ T } has the following
properties? In each case also state one of the following processes which possesses the relevant
property:
White noise, Random walk, Moving average process, Poisson process, Compound
Poisson process, Brownian motion
a) Stationarity
c) Weak Stationarity
d) Independent increments
Question 4 You are thinking of moving to Australia and as part of your investigation of life
on the other side of the ocean you have collected the following sets of data:
1) the maximum daily temperature each day since 1 January 2003
2) whether or not it rained for each day since 1 January 2003
3) the number of cyclists injured in road accidents since 1 January 2003
For each data set choose an appropriate state space and parameter space, indicating for each if
it is discrete or continuous.
1) Parameter Space: each day since 1 Jan 2003, discrete i.e. t = 1,2,3,...
State Space: maximum daily temperature, ℜ : continuous (technically) or (−∞, ∞) : discrete
(more realistic)
2) Parameter Space: each day since 1 Jan 2003, discrete i.e. t = 1,2,3,...
State Space: yes or no i.e. two states, discrete
3) Parameter Space: every possible time since 1 Jan 2003, continuous
State Space: cumulative number of cyclists injured, discrete i.e. S = {0,1,2,...} (counting
process)
WEAKLY STATIONARY: all Yt ’s have the same expected values and is covariance
stationary(i.e. cov(Yt , Ys ) depends only on t − s ).
0 if t > s + 1
2 2
= α 0 σ + α 1 σ 2 if t = s
2
α 0α 1σ if s + 1 = t
2
(1/2 MARK)
which depends only an how far t and s are and not the actual times. (1/2 MARK)
(b) Does this process have the Markov property? Prove your answer.
For any t , Yt = α 0 X t + α 1 X t −1 , and Yt −1 = α 0 X t −1 + α 1 X t − 2 , thus since Yt −1 depends on
X t − 2 , which appears in Yt − 2 , we have that Yt is dependent on at least Yt − 2 and therefore
the Markov Property does not hold. (2 MARK)
(c) Does this process have independent increments? Prove your answer.
Yt +1 − Yt = α 0 X t +1 + α 1 X t − α 0 X t − α 1 X t −1 = α 0 X t +1 + (α 1 − α 0 ) X t − α 1 X t −1 and
Yt −1 = α 0 X t −1 + α 1 X t − 2 thus since the increment depends on X t −1 and Yt −1 depends on X t −1 ,
Yt +1 − Yt does depend on {Ys : 0 ≤ s ≤ t} and we do not have independent increments. (2
MARKS)
Question 6 Consider a branching process with X 0 = 1 i.e. the zeroth generation has only one
element. Let µ be the expected number of offspring of any element and let σ 2 be the variance
of the number of offspring of any element. Let X n be the size of the n th generation. Prove that
σ 2 µ n −1 (1 − µ n )
if µ ≠ 1
E[ X n ] = µ n and var( X n ) = 1− µ .
nσ 2 if µ = 1
Proof
Question 7 a) Define fully a stochastic process, also describing all the elements that make it
up.
d) For a compound Poisson process with parameter λ , prove that the moment generating
function is given by e λt ( M X (u ) −1) .
Proof
b) Give an example of a process which is not stationary. Show that the process given is
indeed not stationary.
The random walk is not stationary:
t
If we let E [X t ] = µ ≠ 0 then E[Yt ] = E ∑ X i = tµ which depends on t . Thus if the
i =1
expected values are not equal the distributions cannot be equal either.
t
c) Consider the random walk process {Yt : t = 0,1,2,...} where Y0 = 0 and Yt = ∑ X i for all
i =1
t ≥ 1 with {X t : t = 0,1,2,...} a white noise process. Show that the random walk possesses the
Markov property.
Question 12 For each of the following processes, provide the state space and parameter space
and state whether they are each discrete, continuous or either. Explain as necessary.
Question 13 Define the terms stationarity, weak stationarity and covariance stationarity.
Also give an example of process(es) which possess these properties and explain why it has
the relevant property.
Question 14 Indicate whether the statements that follow are true or false. If false, explain
why, or correct the statement.
a) To specify the probability structure of a stochastic process {X t : t ∈ T } with the X t ’s
independent, it is required to specify the distribution of each X t for all values of t ∈ T .
True
Question 16 a) Fully define a stochastic process, clearly indicating the spaces involved.
Stationarity
Independent Increments
Stationary Increments
Time-homogeneity
b) For a Poisson process with parameter λ describe which of the properties in (a) hold and
explain why.
c) Derive the moment generating function for a compound Poisson process with parameter λ
. Proof
Question 19 a) Let {Yn : n = 0,1, 2...} be a Markov chain with state space S ⊆ {0,1,2,...}.
i. Show that E (Yn | Y1 , Y2 ,..., Ym ) = E (Yn | Ym ) .
ii. State the two properties that must hold in order to classify this process as a
martingale.
Using these two functions, explain the concept of a deterministic and a stochastic model
respectively.
b) Consider a stochastic process { X t : t ∈ T } . Define and describe the parameter and state
space.
Question 23 Consider a stochastic process { X t : t ∈ T } .
a) Define the property of stationarity.
b) Define the property of covariance stationarity.
c) Prove that a stationary stochastic process is also covariance stationary.
d) Show that a Poisson process with parameter λ is not covariance stationary. Hint: For
t > s, var( N t − N s ) = cov( N t − N s , N t − N s ) = var( N t ) + var( N s ) − 2 cov( N t , N s ) .