Assignment 2 Htop

Name: Thomas Li

Netid: jl15535

Problem 2.1. Show that the assumption that P is countably additive is equivalent to the
assumption that P is continuous. That is to say, show that if a function P : F → [0, 1] satisfies
P(∅) = 0, P(Ω) = 1, and

P(A ∪ B) = P(A) + P(B)

whenever A, B ∈ F and A∩B = ∅, then P is countably additive (in the sense of satisfying Definition
(1.3.1b)) if and only if P is continuous (in the sense of Lemma (1.3.5)).

Solution. Let {An } be a disjoint sequence of sets.

Suppose P is Countable additive. Let {An } be an increasing sequence, define A = An .
S

Then obviously A = A1 ∪ (A2 \A1 ) ∪ (A3 \A2 ) ∪ · · · which is a disjoint union. Then
∞
X
P(A) = P(An+1 \An )
n=1

= P(A1 ) + P(A2 ) − P(A1 ) + · · ·

= lim P(An )
n→∞

Then P is continous.
Suppose P is Continuous. Then pick a sequence {An } of disjoint union. Then define
n
[
Bn = Ai
i=1

Then we see that

∞
[ ∞
[
A= Bn = Ai
n=1 i=1

1
So by the continuity property,
∞ ∞
! !
[ [
P Ai =P Bn
i=1 n=1

= lim P(Bn )
n→∞
n
X
= lim P(Ai )
n→∞
i=1
∞
X
= P(Ai )
i=1

Problem 2.2. The ménages problem poses the following question. Some consider it to be
desirable that men and women alternate when seated at a circular table. If n heterosexual couples
are seated randomly according to this rule, show that the probability that nobody sits next to his
or her partner is
n  
1 X k 2n 2n − k
(−1) (n − k)!
n! 2n − k k
k=0

You may find it useful to show first that the number of ways of selecting k non-overlapping pairs
of adjacent seats is
 
2n − k
2n(2n − k)−1 .
k

Solution. So let the event Ei be i-th man sit next to his wife. Then we want to compute
n
! n
!
\ [
c
P Ei = 1 − P Ei
i=1 i=1

Then using the Inclusiong Exclusion Principle. We need to calculate

X
P(Ei1 ∩ · · · ∩ Eir )
i1 <···ir

with r from 1 to n. We suppose the way of selecting k non-overlapping pairs of adjacent seats is
C(k). Then we have

n C(k)((n − k)!)2 · k! C(k)r!((n − r)!)2

 
X n!
P(Ei1 ∩ · · · ∩ Eir ) = =
i1 <···ir
k ∥Ω∥ r!(n − r)! (n!)2
C(k)(n − r)!
=
n!
So what is left is to compute C(k). So when picking k pairs out of 2n seats. We can mannually
group 2k seats as pairs. Then we would have in total 2n − k objects. Consisting either a single seat

2
or a pair. There are  
2n − k
k
ways of picking k pairs. However, since we assume the selection process having order, we need to
divide be 2n − k for the overcounting. The rotation symmetry will create 2n − 1 more possible
arrangement. Then we have  
2n − k 2n
C(k) =
k 2n − k
So by the Inclusion Exclusion Principle
n
! n
!
\ [
c
P Ei = 1 − P Ei
i=1 i=1
n
1 X
=1− (−1)k−1 C(k)(n − r)!
n!
k=1
n  
1 X k 2n 2n − k
= (−1) (n − k)!
n! 2n − k k
k=0

where we use the fact that  

1 2n 2n
n! = 1
n! 2n − 0 0
□

Problem 2.3. The probabilistic method 10 per cent of the surface of a sphere is coloured
blue, the rest is red. Show that, irrespective of the manner in which the colours are distributed, it
is possible to inscribe a cube in S with all its vertices red.

Solution. Since the placement of the cube is random. Define event Ai as follows.
Let Ai be the event that the i-th vertex lies in a blue region after a random rotation.
Then
P[Ai ] = 0.1
Then consider the event A be At least one Vertex lies in a blue region after a random rotation.Then
S
A ⊂ Ai
X8
P[A] ≤ P[Ai ] = 8 × 0.1 = 0.8
i=1

Then the complement of A is the event that all vertices lie in a red region. So

P[Ac ] ≥ 1 − 0.8 = 0.2

Having a probability of 0.2, it is always possible for the event to happen after finite many trials.

Question: Can we calculate the percentage of covering such that such event arenot meant to happen
for any possible distribution? □

Problem 2.4. Poker. During a game of poker, you are dealt a five-card hand at random. With

3
the convention that aces may count high or low, show that:

P(1 pair) ≃ 0.423, P(2 pairs) ≃ 0.0475, P(3 of a kind) ≃ 0.021,

P(straight) ≃ 0.0039, P(flush) ≃ 0.0020, P(full house) ≃ 0.0014,

P(4 of a kind) ≃ 0.00024, P(straight flush) ≃ 0.000015.

Solution. We calculate the cardinarlity of each event. The cardinarlity of the probability space
Ω is  
52
∥Ω∥ = = 2598960
5
Then

• Event A is 1 pair. Then

∥A∥ = 13 · C42 · 48 · 44 · 40/6
We divide by 6 to eliminate the overcounting since the 3 non-pair cards do not have order.
So
∥A∥
P(A) = ≈ 0.423
∥Ω∥

• Event B is with 2 pair. Then

13 · C42 · 12 · C42 · 44/2

P(B) = ≈ 0.0475
2598960
We divide by 2 because the 2 pairs do not have order.

• Event C is 3 of a kind. Then

13 · C43 · 48 · 44/2
P(C) = ≈ 0.021
∥Ω∥

• Event D is having a straight including straight flush. Then

10 · 45
P(D) = ≈ 0.0039
∥Ω∥

The 45 means that each card in the straight can have four different choices of suits.

• Event E is having a flash. Flash is having all five card in same suit. Then

C1 35 · 4
P(E) = ≈ 0.0020
∥Ω∥

• Event F is having a full house.

13 · C42 · 12 · C42
P(F ) = ≈ 0.0014
∥Ω∥

4
 • Event G is having 4 of a kind.

13 · 48
P(G) = ≈ 0.00024
∥Ω∥

• Event H is having a straight flush. The hardest event.

10 · 4
P(H) = ≈ 0.000015
∥Ω∥

Problem 2.5. Let m be Lebesgue measure on [0, 1], and 0 ≤ a ≤ b ≤ c ≤ d ≤ 1 such that
a + d ≥ b + c. Give an example of a sequence of sets A1 , A2 , . . . in [0, 1], such that

m(lim inf An ) = a, lim inf m(An ) = b, lim sup m(An ) = c, and m(lim sup An ) = d.
n n n n

Solution. Construct in the following way. So for each An there is a fixed part [0, a] union with
a En . Define first that  
d−a
K=
c−a
So we devide the interval [a, d] into K part Ji . Each part having length

d−a
≤c−a
K
So we define a periodic sequence En ⊂ [a, d] consist of:

• K steps covering each Jk with interval of measure c − a.

• 1 step covering a subset of measure b − a within one Jk .

A strekch map for the distribution of En is as follow.

5
 0 a d
J1 J2 J3 J4 J5

c−a

c−a

c−a

c−a

c−a

b−a

We verify the values. Since m(lim inf n En ) =. Then m(lim inf n An ) = a. Since each cycle consist
of a set with length b − a + a = b. Then lim inf n m(An ) = b and similarly, lim supn m(An ) = c.
Also, since each cycle fully covers [a, d] Then m(lim supn An ) = d. The construction is finished.
□

Problem 2.6. Let Ω = R and consider the following subsets of P(R):

C1 := {(−∞, b] : b ∈ R}

C2 := {(a, b] : a, b ∈ R}

C3 := {A ⊂ R, A is closed}.
Show that σ(C1 ) = σ(C2 ) = σ(C3 ).

Solution. First show that σ(C1 ) = σ(C2 ).

Since (a, +∞), (−∞, b] ∈ σ(C1 ). Then (a, b] ∈ σ(C1 ) then C2 ⊂ σ(C1 ). Also pick a = −n then we
see that the countable union
∞
[
(−n, b] = (−∞, b]
n=1

Then C1 ⊂ σ(C2 ).
Then we show that σ(C2 ) = σ(C3 ).
For C3 . Since any closed sets are Gδ set. Since countable union of countable collection remains
countable. So it is enough to construct open set (a, c) with elements in σ(C2 ). Consturct an
increasing sequence {bn } such that bn → c. Then we see that
∞
[
(a, bn ] = (a, c)
n=1

6
Then C3 ⊂ σ(C2 ).
Also, since open sets are Fσ set. Then (a, b) ∈ σ(C3 ). Then take c > a we have

(a, b] = (a, b) ∪ [c, b] ∈ σ(C3 )

then C2 ⊂ σ(C3 ). □