STAT3004/4018/6018 Stochastic Processes, Sem 1
Problem Sheet 4
1 Let ({0, 1}N , P({0, 1})⊗N , Pp ) be the probability space of infinite and independent tossing
a fair coin with success probability p ∈ (0, 1). Here {0, 1}N = {(ω1 ω2 . . . ) : ωk ∈ {0, 1}} are
{0, 1}-sequences, with 0 and 1 encoding head and tail, respectively. The canonical event
system on {0, 1}N is the σ-algebra P({0, 1})⊗N = σ( n∈N {Aa1 ,...,an : a1 , . . . , an ∈ {0, 1}}),
S
generated by cylinder sets of the form Aa1 ,...,an := nk=1 {(ω1 ω2 . . . ) ∈ {0, 1}N : ωk = ak },
T
as associated to n ∈ N and a1 , . . . , an ∈ {0, 1}. The probability measure Pp is determined
by the probabilities of those cylinder sets as
Pn Pn
Pp (Aa1 ,...,an ) = (1−p)n− k=1 ak
p k=1 ak
, n ∈ N, a1 , . . . , an ∈ {0, 1} .
(a) Verify F, G are events where (i) F = {(ω1 ω2 . . . ) : ω1 = 1, ω2 = 0, ω3 = 1)} and
(ii) G =“the outcome is the alternating sequence 010101010 . . . ”.
(b) Determine the probabilities (i) Pp (F ) and (ii) Pp (G) of the events in (a) under the
assumption that the underlying coin is fair.
2 Let (Ω, F, P) be a probability space, and X : Ω → R be a random variable and c ∈ R.
Show that c + X is also a random variable.
3 Let f : R → R be a nondecreasing (=weakly increasing) function, i.e., for all x, y ∈ R,
x ≤ y ⇒ f (x) ≤ f (y). Show that f is a Borel function. Conclude monotonic functions
are Borel functions.
4 Let Ω is a non-empty set, (Ω0 , F 0 ) is a measurable space, X : Ω → Ω0 is a map, and set
σ(X) := {{X ∈ F 0 } : F ∈ F 0 }. Verify
(a) (i) {X ∈ Ω0 } = Ω, (ii) if F 0 ∈ P(Ω0 ), then {X ∈ F 0 }c = {X ∈ F 0c };
(iii) if F10 , F20 , . . . ∈ P(Ω0 ), then ∞
S 0
S∞ 0
n=1 {X ∈ F n } = {X ∈ n=1 Fn };
(b) σ(X) is a σ-algebra on Ω;
(c) if F is a σ-algebra on Ω and, in addition, X is F/F 0 -measurable, then σ(X) ⊆ F.
5? Let Ω, Ω0 be nonempty sets and Z : Ω → Ω0 a map. If F ⊆ P(Ω), let A0 = A0 (Z, F) ⊆
P(Ω0 ) be defined by A0 := A0 (Z, F) := {A0 ⊆ Ω0 : {Z ∈ A0 } ∈ F}. Verify
(a) if T is a non-empty index set and, for t ∈ T , Ft0 ∈ P(Ω0 ), then
n [ o [
Z∈ Ft0 = {Z ∈ Ft0 } ;
t∈T t∈T
(b) if F satisfies (S3) on Ω, then so does A0 on Ω0 ;
(c) if F is a σ-algebra on Ω, then so is A0 on Ω0 .
1
Solutions
1 (a) (i) F ∈ P({0, 1})⊗N . Indeed F = {(ω1 ω2 . . . ) : ω1 = 1, ω2 = 0, ω3 = 1)} = A101 is a
cylinder set (“a tail, a head and a tail in the first, second and third tossing, respectively”).
(ii) G ∈ P({0, 1})⊗N . Indeed, G as a countably infinite intersection of cylinder sets
G = A0 ∩ A01 ∩ A010 ∩ A0101 ∩ . . . .
(b) (i) Assuming a fair coin p = 1/2 yields
1 1
P1/2 (F ) = P1/2 (A101 ) = (1/2)1 (1/2)2 = 3
= .
2 8
(ii) If n ∈ N, introduce the cylinder set Bn := A01010101...01 with the associated length 2n,
so that
∅ ⊆ G ⊆ A01 = B1 , , ∅ ⊆ G ⊆ A0101 = B2 , ... ∅ ⊆ G ⊆ A010101...1 = Bn .
By monotonicity of P1/2 ,
1
0 ≤ P1/2 (G) ≤ P1/2 (A010101...1 ) = P(Bn ) = .
22n
As n → ∞, the RHS in the last display converges to zero, so that P1/2 (G) = 0.
2 Assume c ∈ R is a (deterministic) constant and X : Ω → R is a random variable
(rv). If ω ∈ Ω, set Y (ω) := c + X(ω) ∈ R is well defined real number and ω 7→ Y (ω)
is defined as function from Ω into R. To verify that Y is a rv, it is convenient to use an
interval-generator criterium, i.e. for y ∈ R, we verify {Y ≤ y} ∈ F.
Now X is assumed to be a rv so that, for x ∈ R, {X ≤ x} ∈ F. Now set x := y − c so
that x ∈ R and, as desired,
{Y ≤ y} = {ω ∈ Ω : Y (ω) ≤ y}
= {ω ∈ Ω : c + X(ω) ≤ y}
= {ω ∈ Ω : X(ω) ≤ y − c}
= {X ≤ x} ∈ F.
2
3 Assume a function f : R → R so that, for x ≤ y, x, y ∈ R f (x) ≤ f (y). We verify, for all
z ∈ R, {f ≤ z} = {x ∈ R : f (x) ≤ z} ∈ B = B(R). Pick z ∈ R.
Case 1. If {f ≤ z} ∈ {∅, R}, then {f ≤ z} ∈ {∅, R} ⊆ B. Done.
Case 2. If {f ≤ z} ∈ / {∅, R}, then s := sup{x ∈ R : f (x) ≤ z} is finite real number
s ∈ R. If x ∈ R and x > s, note f (x) > z, implying
(s, ∞) ⊆ {f > z} ⇔ {f ≤ z} ⊆ (−∞, s] .
We have to distinguish 2 subcases. It is recommended to graph the respective subcase.
Subcase 2a. If {f ≤ z} ∈ / {∅, R} and f (s) ≤ z, then {f ≤ z} = (−∞, s] ∈ B (intervals
are Borel sets). Done.
Subcase 2b. If {f ≤ z} ∈ / {∅, R} and f (s) > z, then {f ≤ z} = (−∞, s) ∈ B (intervals
are Borel sets). Done.
Consequently, we found, for z ∈ R, {f ≤ z} ∈ B(R).
Conclusion. Weakly increasing functions are Borel functions. Multiplying those with -1
we see weakly decreasing functions are Borel functions, too. To summarise, monotonic
functions are Borel functions.
4 (a) (i) Since X assigns ω ∈ Ω to ω 0 ∈ Ω0 , the first identity {X ∈ Ω0 } = Ω is plain.
(ii) If F 0 ⊆ Ω0 then, for ω ∈ Ω,
ω ∈ {X ∈ F 0 }c ⇔ not ω ∈ {X ∈ F 0 }
⇔ not X(ω) ∈ F 0
⇔ X(ω) ∈ F 0c
⇔ ω ∈ {X ∈ F 0c } .
(iii) If F10 , F20 , . . . ⊆ Ω0 then, for ω ∈ Ω,
∞
[
ω∈ {X ∈ Fn0 } ⇔ exists n ∈ N, ω ∈ {X ∈ Fn0 }
n=1
⇔ exists n ∈ N, X(ω) ∈ Fn0
∞
[
⇔ X(ω) ∈ Fn0
n=1
n ∞
[ o
⇔ ω∈ X∈ Fn0 .
n=1
3
(b) Recall (Ω0 , F 0 ) is a measurable space, and σ(X) = {{X ∈ F 0 } : F 0 ∈ F 0 } ⊆ P(Ω).
(S1) Noting Ω0 ∈ F 0 and (a)(i) {X ∈ Ω0 } = Ω imply Ω ∈ F.
(S2) Let F ∈ σ(X) so that, for some F 0 ∈ F 0 , F = {X ∈ F 0 }. Being a σ-algebra,
F 0 is closed under taking complement F 0c ∈ F 0 . Now (a)(ii) F c = {X ∈ F 0c } implies
F c = {X ∈ F 0c } ∈ σ(X).
(S3) Let F1 , F2 , . . . ∈ σ(X) be a sequence so that, for n ∈ N, for some Fn0 ∈ F 0 , Fn =
{X ∈ Fn0 }. Being a σ-algebra, F 0 is closed under taking countable unions n∈N Fn0 ∈ F 0 .
S
Now (a)(iii) n∈N Fn = {X ∈ n∈N Fn0 } implies n∈N Fn ∈ σ(X).
S S S
Consequently, {{X ∈ F 0 } : F 0 ∈ F 0 } is a σ-algebra on Ω.
(c) Assume F is a σ-algebra on Ω and X is F/F 0 -measurable map
{X ∈ B} ∈ F ∀B ∈ F 0 .
Pick F ∈ σ(X) so that, for some F 0 ∈ F 0 , F = {X ∈ F 0 }. Thus {X ∈ F 0 } ∈ F. By
arbitrariness of F , this shows the system-inclusion σ(X) ⊆ F.
Conclusion. σ(X) is the smallest σ-algebra on Ω so that X is a measurable map into
the measurable space (Ω0 , F 0 ).
5? Let Ω, Ω0 be nonempty sets and Z : Ω → Ω0 be a map.
(a) If ω ∈ Ω and F 0 ∈ P(Ω0 ), recall ω ∈ {Z ∈ F 0 } ⇔ Z(ω) ∈ F 0 . For ω ∈ Ω,
n [ o [
ω∈ Z∈ Ft0 ⇔ Z(ω) ∈ Ft0
t∈T t∈T
⇔ exists t ∈ T Z(ω) ∈ Ft0
⇔ exists t ∈ T ω ∈ {Z ∈ Ft0 }
[
⇔ ω∈ {Z ∈ Ft0 } .
t∈T
n o S
Consequently, Z ∈ t∈T Ft0 = t∈T {Z ∈ Ft0 }, as desired.
S
4
(b) Let F ⊆ P(Ω), recall A0 = A0 (Z, F) = {A0 ⊆ Ω0 : {Z ∈ A0 } ∈ F} ⊆ P(Ω0 ).
(S3) Let A01 , A02 , . . . ∈ A0 so that, for n ∈ N, Fn := {Z ∈ A0n } ∈ F. Assuming F to
satisfy (S3) on Ω yields ∞
S
n=1 Fn ∈ F. As desired,
n ∞
[ o ∞
[
Z∈ A0n = {Z ∈ A0n } ((a))
n=1 n=1
[∞
= Fn ∈ F ((S3) on Ω)
n=1
[∞
⇒ A0n ∈ A0 ((S3) on Ω0 OK) .
n=1
(c) In view of (b), we are done, provided Axioms (S1) and (S2) hold on (Ω, F) im-
ply Axioms (S1) and (S2) on (Ω0 , A0 ), respectively.
(S1) If Ω ∈ F, then {Z ∈ Ω0 } = Ω ∈ F ⇒ Ω0 ∈ A0 .
(S2) If A0 ∈ A0 , then F := {Z ∈ A0 } ∈ F. Assuming (S2) on Ω yields
(∗)
Fc ∈ F ⇒ {Z ∈ A0c } = {Z ∈ A0 }c = F c ∈ F ⇒ A0c ∈ A0 , (Done) .
We finally verify (∗) by copying and pasting from 4 (a)(ii)), for ω ∈ Ω,
ω ∈ {Z ∈ A0 }c ⇔ not ω ∈ {Z ∈ A0 }
⇔ not Z(ω) ∈ A0
⇔ Z(ω) ∈ A0c
⇔ ω ∈ {Z ∈ A0c } , (Done) .
