STAT3004/4018/6018 Stochastic Processes, Sem 1

Problem Sheet 3

1 If ω ∈ N = {1, 2, . . . }, set p(ω) := (3/4)(1/4)ω−1 . Consider the discrete probability space

(N, P(N), P) where P is given by probability mass function (pmf) ω 7→ P({ω}) = p(ω),
for ω ∈ N.
(a) Verify p : N → [0, 1] is a valid pmf; (b) determine P({2, 3}) and P(2N).
2 Consider the probability space (R, B, P) where B is the Borel σ-algebra and P is deter-
mined by the cumulated distribution function (cdf) ω 7→ F(ω) := I(0,∞) (ω)(1 − e−ω ), for
ω ∈ R. Determine the probabilities (a) P((2, 3]); (b) P((−∞, 1]∪(4, ∞)); (c) P(Q) (pmf).
3 If ∅ 6= Ω0 ⊆ Ω and H ⊆ P(Ω), introduce the trace of system of H with respect to Ω0 ,
i.e. a subsystem H(Ω0 ) ⊆ P(Ω0 ), defined by H(Ω0 ) := {Ω0 ∩ H : H ∈ H}. Verify
(a) if A is an algebra on Ω, then A(Ω0 ) is an algebra on Ω0 ;
(b) if F is a σ-algebra on Ω, then F(Ω0 ) is a σ-algebra on Ω0 ;
(c) if E ⊆ A and A = α(E), then A(Ω0 ) = α0 (E(Ω0 ));
(d) if E ⊆ F and F = σ(E), then F(Ω0 ) = σ 0 (E(Ω0 )).
4 Borel σ-algebra B on R is defined by B := B(R) := σ({(−∞, x] : x ∈ R}). Verify
(a) the identity B = σ({(a, b) : −∞ < a < b < ∞});
(b) the inclusion {{x} : x ∈ R} ⊆ B;
(c) uncountability of B;
(d) existence of a countable π-system E so that B = σ(E);
(e) B([0, 1]) = σ({[0, x] : 0 ≤ x ≤ 1}).
5? Give an example of a finite Ω, a subsystem E and probability measures P1 , P2 , defined
on P(Ω), so that
(a) σ(E) = P(Ω) while, for E ∈ E, P1 (E) = P2 (E) but, for F ⊆ Ω, P1 (F ) 6= P2 (F ).
(b) Determine D := {F ⊆ Ω : P1 (F ) = P2 (F )}.
(c) What went wrong with uniqueness?

1
Solutions
P
1 Geometric distribution. (a) Clearly, p(ω) ∈ [0, 1]. It suffices to show ω∈Ω p(ω) = 1

X ∞
X ∞
X
ω−1
pω = (3/4)(1/4) = (3/4) (1/4)ω
ω∈Ω ω=1 ω=0
∞
X
= (3/4) (1/4)ω = 1 (geometric series) .
ω=0

(b) P({2, 3}) = P({2}) + P({3}) = p(2) + p(3) = (3/4)(1/4)1 + (3/4)(1/4)2 = 15/64 and
∞ ∞
X X
(2ω)−1
X 3 1 1
P(2N) = p(2ω) = (3/4)(1/4) =3 (1/16)ω = 1 = .
ω∈N ω=1 ω=1
16 1 − 16
5


2 Exponential distribution. (a)

P((2, 3]) = P((−∞, 3]) − P((−∞, 2]) = F(3) − F(2) = e−2 − e−3 ≈ 0.085548215 .

(b)

P((−∞, 1]∪(4, ∞)) = P((−∞, 1])+1−P({−∞, 4]}) (finite additivity)

= F(1)+1−F(4) = 1− e−1 + 1− (1 − e−4 ) = 1 − e−1 + e−4 ≈ 0.650436198 .

(c) If x, ε ∈ R and ε > 0, then

0 ≤ P({x}) ≤ P((x − ε, x]) = F(x) − F(x − ε) (monotonicity) .

Our F is a continuous function, as ε ↓ 0,

0 ≤ P({x}) ≤ lim P((x − ε, x]) = 0 .

ε↓0

The probability mass function (pmf) is identically zero, for ω ∈ R, ω 7→ P({ω}) = 0.

Recall Q is countable
X
P(Q) = P({q}) (countable additivity)
q∈Q
X
= 0 = 0.
q∈Q

2
3 (a) Trace algebra. Assume A is an algebra on Ω and ∅ = 6 Ω0 ⊆ Ω. Plainly, A0 :=
A(Ω0 ) := {Ω0 ∩ A : A ∈ A} is a subsystem of P(Ω0 ). We verify A0 is an algebra on Ω0 .
(A1) With A being an algebra on Ω, A obeys (A1) so that Ω ∈ A. Noting Ω0 = Ω0 ∩ Ω
and Ω ∈ A yields Ω0 ∈ A0 .
(A2) Let A0 ∈ A0 so that, for some A ∈ A, A0 = Ω0 ∩ A.
Being an algebra on Ω, A obeys (A2) so that Ac = Ω\A ∈ A. The complement of A0
with respect to Ω0 equals Ω0 \A0 = Ω0 ∩ Ac ⇒ Ω0 \A0 ∈ A0 .
(A3) Let A0 , B 0 ∈ A0 so that, for some A, B ∈ A, A0 = Ω0 ∩ A and B 0 = Ω0 ∩ B.
Being an algebra, A obeys (A3) so that A ∪ B ∈ A. Distributivity implies A0 ∪ B 0 =
Ω0 ∩ (A ∪ B) ⇒ A0 ∪ B 0 ∈ A0 .
(b) Trace σ-algebra. Assume F is a σ-algebra on Ω and ∅ 6= Ω0 ⊆ Ω. Plainly, F 0 :=
F(Ω0 ) := {Ω0 ∩ F : F ∈ F} is a subsystem of P(Ω0 ). We verify F 0 is a σ-algebra on Ω0 .
(S1)=(A1) and (S2)=(A2), see (a).
(S3) Assume F10 , F20 , . . . is a sequence in F 0 so that, for some sequence F1 , F2 , . . . ∈ F,

F10 = Ω0 ∩ F1 , F20 = Ω0 ∩ F2 , . . . , Fn0 = Ω0 ∩ Fn , . . . .

Being a σ-algebra on Ω, F obeys (S3) so that ∞

S S
n=1 Fn = n∈N Fn ∈ F. Distributivity
0 0 0 0
S S
S
implies n∈N Fn = Ω ∩ n∈N Fn ⇒ n∈N Fn ∈ F .

(c) Trace algebra generator. Assume ∅ 6= Ω0 ⊆ Ω, E ⊆ A and A = α(E) on Ω.

Set E 0 := E(Ω0 ) := {Ω0 ∩ E : E ∈ E} so that E 0 ⊆ P(Ω0 ). In (a), we determined
A0 := A(Ω0 ) := {Ω0 ∩ A : A ∈ A} to be an algebra on Ω0 .
Now α(E) = A ⇒ E ⊆ A ⇒ E 0 ⊆ A0 ⇒ α0 (E 0 ) ⊆ A0 , where the last implication holds
because α0 (E 0 ) is the smallest algebra on Ω0 containing E 0 .
Let C 0 be an algebra on Ω0 so that E 0 ⊆ C 0 , and set

C := {C ⊆ Ω : Ω0 ∩ C ∈ C 0 } .

Recall E 0 ⊆ C 0 . If E ∈ E, then Ω0 ∩ E ∈ E 0 ⇒ Ω0 ∩ E ∈ C 0 ⇒ E ∈ C. Thus E ⊆ C. If C is an

algebra on Ω, then E ⊆ C ⇒ A = α(E) ⊆ C ⇒ A0 ⊆ C(Ω0 ) ⊆ C 0 . Thus A0 is the smallest
algebra on Ω0 so that E 0 ⊆ A0 , i.e. A0 = α0 (E 0 ). Still,we must check that C is an algebra
on Ω. Plainly, C ⊆ P(Ω).
(A1) Being an algebra on Ω0 , C 0 obeys (A1) Ω0 ∈ C 0 . Noting Ω0 ∩ Ω = Ω0 yields Ω ∈ C.
(A2) Let C ∈ C so that C 0 := Ω0 ∩ C ∈ C 0 . Being an algebra on Ω0 , C 0 obeys (A2)
Ω0 \C 0 ∈ C 0 . Noting Ω0 ∩ C c = Ω0 \C = Ω0 \C 0 yields C c ∈ C.
(A3) Let C, D ∈ C so that C 0 := Ω0 ∩ C, D0 := Ω0 ∩ D ∈ C 0 . Being an algebra on Ω0 , C 0
obeys (A3) C 0 ∪ D0 ∈ C 0 . Distributivity implies Ω0 ∩ (C ∪ D) = C 0 ∪ D0 ⇒ C ∪ D ∈ C.
(d) Trace σ-algebra generator. Follows by modifying arguments in (b) and (c). 

3
4 (a) Set I := {(−∞, x] : x ∈ R} and C := {(a, b) : −∞ < a < b < ∞}, and verify
I ⊆ σ(C). Pick an arbitrary I ∈ I so that, for some x ∈ R, I = (−∞, x], and observe
[ c
I = (−∞, x] = (x, x + n) ∈ σ(C) .
| {z }
n∈N ∈C

Our identity represents I as complement of a countable union of non-empty, bounded,

open intervals, i.e. elements of C. We conclude I ∈ σ(C). Thus I ⊆ σ(C) is verified.
C ⊆ σ(I). Pick an arbitrary C ∈ C so that, for some a, b ∈ R a < b, C = (a, b), and
observe  c [ 
C = (a, b) = (−∞, a] ∩ − ∞, b − (1/n) ∈ σ(I) .
| {z } | {z }
∈I n∈N
∈I

Our identity represents C as pairwise intersection of a complement and a countable union

of elements of I. We conclude C ∈ σ(I). Thus C ⊆ σ(I) is verified.
As I ⊆ σ(C) and C ⊆ σ(I) holds, (a) follows from monotonicity and idempotence of σ

B := σ(I) ⊆ σ(σ(C)) = σ(C)) ⊆ σ(σ(I)) = σ(I) = B ⇒ B = σ(I) = σ(C) .

(b) Pick x ∈ R, and verify the singleton {x} is a Borel set, i.e. {x} ∈ B,
 c [ 
{x} = (−∞, x] ∩ − ∞, x − (1/n) ∈ σ(I) = B .
| {z } | {z }
∈I n∈N
∈I

Our identity represents {x} as pairwise intersection of a complement and a countable

union of elements of I.
(c) {x} 7→ x is a bijection from the system X := {{x} : x ∈ R} into the uncountable
set R. As a result, X is an uncountable system. In (b), we verified X ⊆ B, and so B is
must be an uncountable system, too.
(d) Recall the set Q of rational numbers, and set

Q := (−∞, r] : r ∈ Q .

Q is countable system. Note Q is a countable set while (−∞, r] 7→ r is a bijection from

the system Q into the countable set Q, implying countability of Q.
Q is a π-system. Note 0 ∈ Q ⇔ (∞, 0] ∈ Q ⇒ Q = 6 ∅. Further, Q is closed under pairwise
intersections; pick arbitrary Q, R ∈ Q so that, for some q, r ∈ Q, Q = (−∞, q] and
R = (−∞, r]. Next, note q ∧ r = min{q, r} ∈ Q, such that
i
Q ∩ R = (−∞, q] ∩ (−∞, r] = 0, r ∧ q ∈ Q (Done) .
| {z }
∈Q

4
σ(Q) = B. Note Q ⊆ I so that σ(Q) ⊆ σ(I) = B. To see B = σ(I) ⊆ σ(Q), we verify
I ⊆ σ(Q). Let I ∈ I so that, for some x ∈ R, I = (−∞, x], and note Q ∩ (x, ∞) is a
countably infinite set while
\
I = (−∞, x] = (−∞, r] ∈ σ(C)
| {z }
r∈Q∩(x,∞) ∈Q

Our identity represents I as a countable intersection of elements of Q. Without proof

we use the fact that any finite real number can be approximated from above by rational
numbers.
(e) This follows by inspecting Parts (b) and (d) of Question 3, e.g. Ω0 := [0, 1] ⊆ R =: Ω,
E := I = {(−∞, x] : x ∈ R}, B = σ(I) etc. 
5? (a) Set Ω := {1, 2, 3, 4} and E := {{1, 2}, {1, 3}}. Note |Ω| = 4. Thus α(E) = σ(E).
Successively, argue as follows
{1} = {1, 2}∩{1, 3} ∈ σ(E) = α(E) as (σ-)algebras are closed under pairwise intersections;
{2} = {1, 2}\{1} ∈ σ(E) = α(E) and {3} = {1, 3}\{1} ∈ σ(E) = α(E) as (σ-)algebras are
closed under set-theoretical differences.
Since Ω = {1, 2, 3, 4} is a finite set and {{1}, {2}, {3}} ⊆ α(E) = σ(E) is finite we have
α(E) = σ(E) = P(Ω) (discrete case where subsets are events, and vice versa) .
Since we are in the discrete setting, i.e countable Ω and F = P(Ω)), it is convenient to
define 2 distinctive probability measures P1 , P2 : P(Ω) → [0, 1] via 2 distinctive probability
mass functions (pmfs) p1 , p2 : Ω → [0, 1], respectively,
ω 7→ p1 (ω) = P1 ({ω}) = const = 1/4 ⇐⇒ (Ω, P(Ω), P1 ) is the Laplace experiment.
( )
1/3 if ω ∈ {1, 4}
ω 7→ p2 (ω) = ⇐⇒ (Ω, P(Ω), P2 ) is not the Laplace experiment.
1/6 if ω ∈ {2, 3}
Let P2 : P(Ω) → [0, 1] be specified as P2 ({1}) = P2 ({4}) = 1/3 and P2 ({2}) =
P2 ({3}) = 1/6 such that (Ω, P(Ω), P2 ) is not the Laplace experiment.
Plainly, P1 6= P2 ⇔ p1 6= p2 .
If E ∈ E, we have P1 (E) = P2 (E)
2 1 1 1
P1 ({1, 2}) = = = + = p2 (1) + p2 (2) = P2 ({1}) + P2 ({2}) = P2 ({1, 2})
4 2 3 6
and
2 1 1 1
P1 ({1, 3}) = = = + = p2 (1) + p2 (3) = P2 ({1}) + P2 ({3}) = P2 ({1, 3}) .
4 2 3 6
(b) D = {∅, {1, 2}, {1, 3}, {3, 4}, {2, 4}, {1, 2, 3, 4}} =
6 P(Ω) = σ(E). (‘D’ refers to Dynkin
system aka λ-system.)
(c) E is not a π-system, not closed under pairwise intersection {1, 2} ∩ {1, 3} = {1} ∈
/ E.