AU7022 Stochastic Methods in Systems & Control Xiang Yin

1 Probability Space

Some Basic Knowledge about Set Theory

▶ A set A is finite if it has finite elements, i.e., A = {a1 , a2 , . . . , an }; otherwise, it is infinite.

▶ A set A is countable if it has a bijection to a subset of N, i.e., it can be “listed”; otherwise

it is uncountable.

▶ A finite set is always countable. If a set is countably infinite, then it is in the form of
A = {a1 , a2 , . . . , an , . . . }. Real numbers R or any interval [a, b] ⊆ R is uncountable.

▶ The power set of A is the collection of all its subsets, i.e., 2A = {B : B ⊆ A}.

Motivating Examples
▶ Example 1: Experiment with Finite Outcomes

– Consider a simple experiment: a coin is tossed twice.

– There are four possible outcomes: HH, HT, T H, T T
– We can talk about events like: A =“no H appears”, B =“T appears at least once”,
C=“T appears no less than H” ...
– Each of the above events has a probability: P (A) = 14 , P (B) = P (C) = 34 . Actually,
B and C are the same event.

▶ Example 2: Experiment with Countably Infinite Outcomes

– Consider another experiment: keep tossing a coin until H appears.

– There are countably infinite possible outcomes: H, T H, T T H, . . .
– We can talk about events like:
Ak =“H appears exactly in the kth toss” or Bk =“H appears in the first k tosses”.
– Then we have probability as follows: ∑
P (A1 ) = 12 , P (A2 ) = 14 , P (A3 ) = 18 , . . . , → 0, which gives ∞
k=1 P (Ak ) = 1; and
P (B1 ) = 2 , P (B2 ) = 4 , P (B3 ) = 8 , . . . , → 1.
1 3 7

▶ Example 3: Experiment with Uncountable Outcomes

– Consider the experiment: randomly take a single point in a disk of unit radius.
– There are uncountably many possible outcomes: all points in the unit disk.
– We can talk about events like:
A =“the point is on the boundary”, B =“the distance to 0 is between 0.3 and 0.5”.
– Then we can assign probability: P (A) = 0 and P (B) = 0.25 − 0.09 = 0.16.

1
AU7022 Stochastic Methods in Systems & Control Xiang Yin

An Informal Summary
We can conclude the followings from the above example:

▶ The most basic thing in an experiment is the sample space Ω, which is the set of all
outcomes. For example, Ω = {HH, HT, T H, T T } for Example 1, Ω = {H, T H, T T H, . . . }
for Example 2 and Ω = {(x, y) ∈ R2 : x2 + y 2 ≤ 1} for Example 3.

▶ An event is a subset of the sample space Ω, e.g., A = {HT, T H, T T }, B =

{H, T H, T T H, T T T H} or C = {(x, y) ∈ R2 : x2 + y 2 = 1}.

▶ An event A occurs if the outcome ω of the experiment is in A, i.e., ω ∈ A.

▶ A probability function that assigns each event a probability.

What We Need for Events

▶ If A and B are events, then we can think A ∪ B, A ∩ B and Ac as new events
“A or B”, “A and B” and “no A”, respectively.

▶ Events A and B are called disjoint if A ∩ B = ∅.

▶ The empty set ∅ is the impossible event and the set Ω is the certain event.

▶ However, not an arbitrary subset of the sample space needs to be an event; we only want
consider those events in whose occurrences we may be interested. In fact, we may have
problem if we do not define events carefully!

▶ Let F be the collection of all events. Apparently we have F ⊆ 2Ω , but what else do we
need? The answer is σ-field defined as follows:

Definition: σ-Field

Let F ⊆ 2Ω be a set of subsets of Ω. We say that F is a σ-field on Ω if it satisfies the

following conditions:

1. ∅ ∈ F;

2. if A ∈ F , then Ac ∈ F ;
∪∞
3. if A1 , A2 , · · · ∈ F , then i=1 Ai ∈ F .

• For example, for any Ω, F = {∅, Ω}, F = {∅, A, Ac , Ω} and F = 2Ω are all σ-fields.

• For σ-field F, if A, B ∈ F , then A ∩ B, A \ B ∈ F .

Proof: A ∩ B = (Ac ∪ B c )c and A \ B = A ∩ B c .

• If F1 , F2 are σ-fields, then F1 ∩ F2 is also a σ-field. (Try to proof as a homework)

• However, F1 ∪ F2 may not be a σ-field. Consider F1 = {∅, A, Ac , Ω}, F2 = {∅, B, B c , Ω}.

• Question: How F looks like if Ω = R or Ω = [a, b]?

2
AU7022 Stochastic Methods in Systems & Control Xiang Yin

What We Need for Probability

▶ Suppose that we perform the experiment for N times and denote by N (A) the number
of times event A occurs, then we have P (A) ≈ NN(A) .

▶ Clearly N (∅) = 0 and N (Ω) = N . Furthermore, if A and B are disjoint events, then
N (A ∪ B) = N (A) ∪ N (B). This suggests that P should be countably additive.

Definition: Probability Measure and Probability Space

Let Ω be a sample space and F ⊆ 2Ω be a σ-filed on Ω. A function P : F → [0, 1] is

said to be a probability measure on (Ω, F) if it satisfies the following conditions:
1. P (∅) = 0, P (Ω) = 1;
2. for any events A1 , A2 , · · · ∈ F such that ∀i ̸= j : Ai ∩ Aj = ∅, we have
(∞ )
∪ ∑ ∞
P Ai = P (Ai )
i=1 i=1

Then 3-tuple (Ω, F, P ) is said to be a probability space if F ⊆ 2Ω is a σ-filed on Ω

and P is a probability measure on (Ω, F).

• Remark (Atom)
Sometimes it is not convenient to list all events in F. But since F is a σ-field, if A, B ∈ F ,
then we must have A ∪ B ∈ F . Therefore, it suffices to list A and B. Given (Ω, F, P ),
we say A ∈ F is an atom if P (A) > 0 and ∀B ⊂ A : P (B) < P (A) ⇒ P (B) = 0. For
example, for F = {∅, Ω, A, B, C, A ∪ B, B ∪ C, A ∪ C}, atoms are A, B and C.

Properties of Probability Space

▶ For any A ∈ F , we have P (Ac ) = 1 − P (A).

Proof: (i) P (A ∪ Ac ) = P (Ω) = 1; (ii) P (A ∪ Ac ) = P (A) + P (Ac ) since A ∩ Ac = ∅.

▶ If A1 , A2 ∈ F and A1 ⊆ A2 , then P (A2 ) = P (A1 ) + P (A2 \ A1 ) ≥ P (A1 ).

Proof: Because A1 and A2 \ A1 are disjoint.

▶ For A1 , A2 ∈ F , we have P (A1 ∪ A2 ) = P (A1 ) + P (A2 ) − P (A1 ∩ A2 ).

Proof: P (A1 ∪ A2 ) = P (A1 ∪ (A2 \ A1 )) = P (A1 )+ P (A2 \ A1 ) = P (A1 )+ P (A2 \ (A1 ∩ A2 ))
▶ For A1 , A2 . . . , An ∈ F , we have
( )
∪
n ∑ ∑ ∑
P Ai = P (Ai )− P (Ai ∩Aj )+ P (Ai ∩Aj ∩Ak )−· · ·+(−1)n+1 P (A1 ∩A2 ∩· · ·∩An )
i=1 i i<j i<j<k

Proof: Leave as a homework (Hint: by induction)

▶ For any sequence of events A1 , A2 , · · · ∈ F , we have

(∞ ) ( ) ( n )
∪ ∪n ∪
P Ak = P lim Ak = lim P Ak ; the same for ∩
n→∞ n→∞
k=1 k=1 k=1

3
AU7022 Stochastic Methods in Systems & Control Xiang Yin

Conditional Probability
▶ Suppose that we perform an experiment N times. Then the conditional probability
for “if B occurs, then the probability of A is p” can be counted as NN(A∩B)
(B)
= NN(A∩B)/N
(B)/N
≈
P (A∩B)
P (B)
. This motivates the define conditional probability.

Definition: Conditional Probability

If P (B) > 0, then the conditional probability that A occurs given that B occurs is

P (A ∩ B)
P (A | B) =
P (B)

P (A) is the prior probability of A; P (A | B) is the posterior probability of A given B.

• Law of Total Probability & Bayes’ Rule

Let A1 ∪A
˙ 2 ∪˙ . . . ∪A
˙ n = Ω be a partition of Ω and Ai ∈ F . Then for any B ∈ F , we have

∑
n ∑
n
P (B) = P (B ∩ Ω) = P (B ∩ (A1 ∪ A2 ∪ . . . An )) = P (B ∩ Ai ) = P (B | Ai )P (Ai )
i=1 i=1

This also leads to the well-known the Bayes’ Rule

P (B | Ai )P (Ai ) P (B | A)P (A)

P (Ai | B) = ∑n or P (A | B) =
i=1 P (B | Ai )P (Ai ) P (B | A)P (A) + P (B | Ac )P (Ac )

Independence

▶ In general P (A | B) changes when B changes unless A does not depend on B. Let us

consider the follows equivalence
P (A ∩ B) P (A ∩ B c ) P (A ∩ B) P (A) − P (A ∩ B)
P (A | B) = P (A | B c ) ⇔ = ⇔ = ⇔ P (A ∩ B) = P (A)P (B)
P (B) P (B c ) P (B) 1 − P (B)

Definition: Independence

We say two events A, B ∈ F are (statistically) independent if P (A ∩ B) = P (A)P (B).

A set of events A1 , A2 , . . . , An ∈ F are said to be mutually independent if

∏
l
∀{k1 , k2 , . . . , kl } ⊆ {1, 2, . . . , n} : P (Ak1 ∩ Ak2 ∩ · · · ∩ Akl ) = P (Aki )
i=1

• Remark: Pairwise independence does not imply mutual independence

Let Ω = {1, 2, . . . , 7}, F = 2Ω , P ({1}) = P ({2}) = · · · = P ({6}) = 18 , P ({7}) = 1
4
.
Consider A = {1, 2, 7}, B = {3, 4, 7}, C = {5, 6, 7}, so P (A) = P (B) = P (C) = 1
2
.
However, these three events are
(1) pairwise independent, since P (A ∩ B) = P (B ∩ C) = P (A ∩ C) = 41 ;
(2) not mutually independent, since P (A ∩ B ∩ C) = 1
4
̸= P (A)P (B)P (C) = 18 .

4
AU7022 Stochastic Methods in Systems & Control Xiang Yin

Examples for Conditional Probability & Independence

▶ Example 1
There are two kinds of COVID-19 vaccines distributed randomly in Brazil:

– 70% are the Chinese vaccines and 30% are the US vaccines.
– The effectiveness of the Chinese vaccine is 90% and that of the US vaccine is 50%.

The experiment is that a person takes a vaccine randomly and waits for the effectiveness.

Then the sample space is Ω = {C+, C−, U +, U −}.

Let A = {C+, U +} be the event that he becomes immune and B = {U +, U −} be the

event that he takes the US vaccine. Then we have

P (A) = P (A | B)P (B) + P (A | B c )P (B c ) = 0.5 × 0.3 + 0.9 × 0.7 = 0.78

If the vaccine fails, then the probability that he took the US vaccine is

P (B ∩ Ac ) 0.3 × 0.5
P (B | Ac ) = = = 68.2%
c
P (A ) 1 − 0.78

Since P (A ∩ B) = 0.5 and P (A) = 0.78, events A and B are clearly dependent.

▶ Example 2
Choose a card randomly from a pack of 52 cards. Then we have
4 1 1
P (A) = , P (♣) = and P (♣A) =
52 4 52
So we conclude that the suit of the choose card is independent of its rank.