TU München Probability Theory

Zentrum Mathematik Summer Term 2013/14

Lehrstuhl für Wahrscheinlichkeitstheorie Berger/Salvi/Miller

Solutions for Exercise sheet 1

σ-Algebras, Measurable Maps and Random Variables
Solution for Exercise 1.1

(i) The identity is measurable.

(ii) Take the map X : Ω1 → Ω2 such that X(1) = 1, X(2) = 1 and X(3) = 2. X is ot
measurable, since the pre-image of the set {1} ∈ F2 is the set {1, 2} 6∈ F1 .
(iii) There are 33 = 27 different maps between Ω1 and Ω2 (for each of the 3 elements in
Ω1 there are 3 possibile images in Ω2 ). Only 15 of them are measurable. Keeping
X(1) = 1, there are five possibilities for the images of 2 and 3:
• 2→1 and 3 → 1;
• 2→2 and 3 → 3;
• 2→3 and 3 → 2;
• 2→2 and 3 → 2;
• 2→3 and 3 → 3.
The same holds when we keep X(1) = 2 or X(1) = 3, for a total of 15 combina-
tions. All the other maps are not measurable (see e.g., part (ii)).

Solution for Exercise 1.2

(i) If X and Y are random variables, then also Z = X + Y is a random variable.

In fact, we only have to verify (see Hint) that {ω ∈ Ω : X(ω) + Y (ω) < q} ∈ F
Q
for all q ∈ . We can rewrite
{X + Y < q} = {X < q − Y }
[

= {X < a} ∩ {a < q − Y } .
a∈ Q
Q
Since {X < a} and {a < q − Y } are in F for each a ∈ (by measurability of X
and Y ), also the countable union of the intersection of such events is in F.
(ii) If X + Y is a random variable, X and Y do not have necessarily to be random
variables. As an easy counterexample take Ω = {ω1 , ω2 , ω3 , ω4 } equipped with
the σ-algebra F = {Ω, {ω1 , ω2 }, {ω3 , ω4 }, ∅} and the following random variables:
X : Ω → R such that X(ω1 ) = 1, X(ω2 ) = 2, X(ω3 ) = 1, X(ω4 ) = 2, and
Y : Ω → R such that Y (ω1 ) = −1, Y (ω2 ) = −2, Y (ω3 ) = −1, Y (ω4 ) = −2, .
Then neither X nor Y is a random variable (e.g., X −1 (2) = {ω2 , ω4 } 6∈ F), but
X + Y ≡ 0 is constant, hence measurable.
Solution for Exercise 1.3
The first version of the exercise was misunderstandable, sorry for that. |A| could be
intended as the cardinality of A or as its Lebesgue measure. This does not make a
big difference for the purposes of part (i), but can make things trickyer for part (ii).
Apologies.
(i) F is an algebra:
– Ω= R is in F, since Ωc = ∅ is clearly finite.
– If A, B ∈ F and are finite, then also A ∩ B is finite. If, w.l.o.g., Ac and B are
finite, then A∩B is finite. If both Ac and B c are finite, then (A∩B)c = Ac ∪B c
is also finite.
– If A ∈ F is finite, then (Ac )c is of course finite, so that Ac ∈ F, too. If on the
other hand A ∈ F and Ac is finite, then it is clear that Ac ∈ F.
BUT F is not a σ-algebra! Take for example A1 ,S
A2 , ... such that Ai = {i} for all
N
i ∈ . Then clearly Ai ∈ F for all i´s, but A = ∞i=1 Ai is not: both A = and N
c
R N
A = \ are not finite.
(ii) We have to enlarge F to the smallest σ-algebra containing F. Consider
F̃ = {A ⊂ R : either A is countable or Ac is countable}.
We claim that F̃ = σ(F).
– σ(F) ⊆ F̃ : since clearly F ⊂ F̃ , all we have to do is to show that F̃ is a
σ-algebra.
∗ R ∈ F̃ is again ok.
∗ If A, B ∈ F̃ and are countable, then also A ∩ B is countable. If, w.l.o.g.,
Ac and B are countable, then A ∩ B is countable. If both Ac and B c are
countable, then (A ∩ B)c = Ac ∪ B c is also countable.
∗ If A ∈ F̃ is countable, then (Ac )c is of course countable, so that Ac ∈ F̃ ,
too. If on the other hand A ∈ F̃ and Ac is countable, then it is clear that
Ac ∈ F̃ .
∗ Finally, the countable union of countable sets is again countable.
– F̃ ⊆ σ(F): this is maybe the hardest part. Let A ∈ F̃ be countable (otherwise
we can apply the same way of reasoning to Ac ). Then there exists a bijection b
N
between A and . Let S
∞
N
Ak := b−1 ({1, 2, ..., k}) for all k ∈ . Note
S∞ that each Ak
is in F and that A = i=1 Ai . Since, by definition of σ(F), i=1 Ai ∈ σ(F),
we have that also A ∈ σ(F).

Solution for Exercise 1.4

(i) Clearly A ⊆ Aµ , since every element of A is also in Aµ (take E = F = A in the

definition of Aµ ).
We have to prove the four σ-algebra properties:
 (i) Ω ∈ Aµ is clear (take E = F = Ω in the definition of Aµ ).
(ii) Take A1 and A2 in Aµ . We have to prove that A1 ∩ A2 ∈ Aµ .
We know that there exist E1 , E2 , F1 , F2 ∈ A such that: Ei ⊆ Ai ⊆ Fi and
µ(Fi \ Ei ) = 0 for i = 1, 2. First observe that E1 ∩ E2 ∈ A and F1 ∩ F2 ∈ A.
Also, by De Morgan laws, (F1 ∩ F2 ) \ (E1 ∩ E2 ) ⊆ (F1 \ E1 ) ∪ (F2 \ E2 ) so that
the l.h.s. has zero µ measure, being a subset of the union of two zero-measure
sets. Now, it easy to see that E1 ∩ E2 ⊆ A1 ∩ A2 ⊆ F1 ∩ F2 , so we are done.
(iii) If A ∈ Aµ , with E and F as before, A 3 F c ⊆ Ac ⊆ E c ∈ A and µ(E c \ F c ) =
0 (if this was not the case, than there would exist C ⊆ E c \F c with µ(C) > 0,
and we would arrive to the contradiction µ(F \ E) ≥ µ(C) > 0).
(iv) For proving the property for infite intersections, just repeat the argument of
part (ii) for infitely many Ai ´s.

(ii) µ̄(A) = µ(A) for A ∈ A is clearly true by the definition of µ̄. We are left to show
that µ̄ is indeed a measure on Aµ .

• µ̄(∅) = 0 is again clear;

• Now take A1 , A2 , ... ∈ Aµ disjoint sets. Let E1 , E2 , ... ∈ A and F1 , F2 , ... ∈ A
be such that Ei ⊆ Ai ⊆ Fi and µ(Fi \ Ei ) = 0 (i.e., µ(Ei ) = µ(Fi )). Then it
follows that µ(Ei ) = µ̄(Ei ) ≤ µ̄(Ai ) ≤ µ̄(Fi ) = µ(Fi ). At this point it easy to
see that, since also the Ei ´s have to be disjoint,
∞
X ∞
X ∞
[  ∞
[ 
µ̄(Ai ) = µ̄(Ei ) = µ̄ Ei = µ̄ Ai .
i=1 i=1 i=1 i=1