PROBABILITY AND STOCHASTIC PROCESS (MTL106)

ANANTA KUMAR MAJEE

1. Probability Theory:
Consider the experiment of tossing a coin. There are two possible outcome {H, T } and we
don’t know whether H or T will come in advance for any performance of this experiment. From
this example, we arrive at the following:
Definition 1.1 (Random Experiment). A random experiment is an experiment in which
i) all outcomes of the experiment are known in advance
ii) any performance of the experiment results in an outcome that is not known in advance
iii) the experiment can be repeated under identical conditions.
Definition 1.2 (Sample space). The set Ω of all possible outcomes of a random experiment
is called sample space. An element ω ∈ Ω is called a sample point.
A die is thrown once. We are interested in the possible occurrences of the ’events’
a) the outcome is the number 1
b) the outcome is even number
c) the outcome is even but does not exceed 3
d) the outcome is not even
Each of the above events can be specified as a subset A of the sample space Ω = {1, 2, 3, 4, 5, 6}
as follows:
a) A = {1}, b) A = {2, 4, 6}, c) A = {2, 4, 6} ∩ {1, 2, 3}, d) A = {2, 4, 6}∁
Thus, events are subsets of Ω, but need all the subsets of Ω be events? The answer is NO. We
need the collection of events be closed under complement and countable union.
Definition 1.3 (σ-algebra). Let Ω ̸= ∅. A collection F of subsets of Ω is called a σ-algebra
or σ-field over Ω if the following properties hold:
i) ∅ ∈ F
ii) If A1 , A2 , . . . ∈ F, then ∪∞
i=1 Ai ∈ F
iii) If A ∈ F, then A . ∁

The elements of F are called ‘events’. One can easily show that if A, B ∈ F, then A ∩ B ∈ F.
Example 1.1. i) The smallest σ-field associated to Ω is the collection F = {∅, Ω}.
ii) If A is any subset of Ω, then F = {∅, A, A∁ , Ω} is a σ-field.
ii) P(Ω) := {A : A ⊆ Ω} is a σ-field, called the total σ-field over Ω.
Remark 1.1. In general, union of σ-algebras over Ω may not be a σ-algebra. For example, let
Ω = {1, 2, 3} and F1 = {∅, {1}, {2, 3}, Ω} and F2 = {∅, {1, 2}, {3}, Ω}. Then both F1 and F2 are
σ-algebra, but F1 ∪ F2 is NOT a σ-algebra.
Lemma 1.1. Intersection of two σ-fields over Ω is a σ-field over Ω. More generally, if {Fi :
i ∈ I} is a family of σ-fields over Ω, then G = ∩i∈I Fi is also a σ-field.
1
2 A. K. MAJEE

Definition 1.4 (Generated σ-field). Let Ω ̸= ∅ and A be be collection of subsets of Ω. Let

M := {F : F is a σ-field over Ω containing A}.
Then σ(A) := ∩F ∈M F is the smallest σ-field over Ω containing A. σ(A) is called the σ-field
generated by A.
Example 1.2 (Borel σ-field). The smallest σ-algebra over R containing all intervals of the
form (−∞, a] with a ∈ R is called the Borel σ-algebra and denoted by B(R).
• Any A ∈ B(R) is called a Borel subset of R.
• The followings are Borel subsets of R: for a ≤ b
1
(a, ∞) = R \ (−∞, a], (a, b] = (−∞, b] ∩ (a, ∞), (∞, a) = ∪∞ n=1 (−∞, a − ],
 n 
[a, ∞) = R \ (−∞, a), (a, b) = (−∞, b) ∩ (a, ∞), [a, b] = R \ (−∞, a) ∪ (b, ∞) ,
m
{a} = [a, a], N = ∪∞ n=0 {n}, Q = ∪m,n∈Z,n̸=0 { }, Q∁ = R \ Q.
n
Example 1.3 (Borel n
Q σ-algebra on R ). It is σ-algebra generated by the n-dimensional rect-
angles of the form ni=1 (ai , bi ]
Definition 1.5 (Measurable space). Let Ω ̸= ∅ and F be a sigma-algebra over Ω. The pair
(Ω, F) is called a measurable space.
Our goal now is to assign to each event A, a non-negative real number which indicates its
chance of happening.
Definition 1.6 (Probability measure). Let (Ω, F) is a measurable space. A function P :
F → [0, 1] is called probability measure if
i) P(A) ≥ 0 ∀ A ∈ F
ii) P(∅) = 0 and P(Ω) = 1
iii) If {Ai } ⊂ F and Ai ’s are pairwise disjoint, then
∞

X
P ∪∞
i=1 A i = P(Ai ).
i=1
The triplet (Ω, F, P) is called a probability space.
Example 1.4. Let Ω = {1, 2, 3} and F = {∅, Ω, {1}, {2, 3}}. Define P : F → [0, 1] via
(
1, if 3 ∈ A
P(A) =
0, if 3 ∈ / A.
Then P is a probability measure on (Ω, F).
Example 1.5. Let Ω = (0, ∞) and F is the Borel σ-field on Ω. Define, for each interval A ⊂ Ω
Z
P(A) = e−x dx.
A
R −x
Observe that P(A) ≥ 0 for all A ∈ F and Ω e dx = 1. Let {Ai } ⊂ B and pairwise disjoint.
∞

R −x
Then P ∪i=1 Ai = R e 1∪∞ i=1 Ai
(x) dx. We need to show that
Z ∞ Z
X ∞
X
−x −x
e 1∪∞ i=1 Ai
(x) dx = e dx = P(Ai ).
R i=1 Ai i=1
One can show the first equality by using monotone convergence theorem on the sequence of
functions fn (x) = e−x 1∪ni=1 Ai (x).
 PROBABILITY AND STOCHASTIC PROCESS 3

Example 1.6. Let Ω = {1, 2, 3, . . .} and F be the set of all subsets of Ω. Define P : F → [0, 1]
via
1
P({i}) = i .
2
Note that since ∞ 1
P
i=1 2i = 1, we have P(Ω) = 1. Check that P is a probability measure on (Ω, F).

Theorem 1.2. Let (Ω, F, P) be a probability space. Then the followings hold:
1) P(A∁ ) = 1 − P(A) for any A ∈ F.
2) If A ⊆ B, then P(A) ≤ P(A) + P(B \ A) = P(B).
3) P(A ∪ B) = P(A) + P(B) − P(A ∩ B). More generally, if A1 , A2 , . . . , An are events, then
n
X X X
P(∪ni=1 Ai ) = P(Ai ) − P(Ai ∩ Aj ) + P(Ai ∩ Aj ∩ Ak ) − . . . + (−1)n+1 P(∩ni=1 Ai ).
i=1 i<j i<j<k

4) Let {An } be an increasing sequence of elements in F i.e., An ∈ F and An ⊆ An+1 for

all n = 1, 2, . . .. Then
P(∪∞
n=1 An ) = lim P(An ). (1.1)
n→∞

5) Let {An } be a decreasing sequence of elements in F i.e., An ∈ F and An+1 ⊆ An for all
n = 1, 2, . . .. Then
P(∩∞
n=1 An ) = lim P(An ). (1.2)
n→∞

The propery (1.1) is called continuity of P from below and property (1.2) is called continuity of
P from above.
Proof. Proof of 1): Since A ∪ A∁ = Ω, we see that P(A∁ ) = 1 − P(A) for any A ∈ F.
Proof of 2): Observe that
B = A ⊔ (B \ A) =⇒ P(B) = P(A) + P(B \ A) ≥ P(A).
Proof of 3): We have
A ∪ B = A ⊔ (B \ A) =⇒ P(A ∪ B) = A + P(B \ A). (1.3)
Observe that B \ A ⊆ B, and hence by 2), we get that


P(B) = P(B \ A) + P B \ (B \ A) = P(B \ A) + P(A ∩ B)
=⇒ P(B \ A) = P(B) − P(A ∩ B). (1.4)
Combining (1.3) and (1.4), we obtain
P(A ∪ B) = P(A) + P(B) − P(A ∩ B).
Proof of 4): Let A = ∪∞
n=1 Ai . Define

C 1 = A1 , C2 = A2 \ A1 , C3 = A3 \ A2 , . . . , Cn = An \ An−1 .
Then {Cn } are pairwise disjoint and A = ∪∞ n
n=1 Cn with An = ⊔i=1 Ci . Hence we have
∞
X n
X
P(A) = P(∪∞
n=1 Cn ) = P(Cn ) = lim P(Ck ) = lim P(∪nk=1 Ck ) = lim P(An ).
n→∞ n→∞ n→∞
n=1 k=1

Proof of 5): Since An ’s are deceasing, A∁n are increasing. Hence by 4), we get
P(∪∞ ∁ ∁ ∞ ∁ ∁
= lim P(An ) =⇒ P(∩∞


n=1 An ) = lim P(An ) =⇒ P ∪n=1 An )
n→∞ n→∞ n=1 An ) = lim P(An ). n→∞
□
4 A. K. MAJEE

Example 1.7. Let (Ω, F, P) be a probability space and let A, B ∈ F such that P(A) = 43 and
P(B) = 13 . Show that
1 1 3
≤ P(A ∩ B) ≤ , ≤ P(A ∪ B) ≤ 1.
12 3 4
Solution: Since P(A ∩ B) = P(A) + P(B) − P(A ∪ B), and P(A ∪ B) ≤ 1, we have
1 3 1
P(A ∩ B) ≥ −1 + + = .
3 4 12
Also, since A ∩ B ⊆ A and A ∩ B ⊆ B, we have
1
P(A ∩ B) ≤ min{P(A), P(B)} = .
3
Since A ⊆ A ∪ B and B ⊆ A ∪ B, and fact that P(A ∪ B) = P(A) + P(B) − P(A ∩ B), we get
max{P(A), P(B)} ≤ P(A ∪ B) ≤ min{P(A) + P(B), 1}
3
=⇒ ≤ P(A ∪ B) ≤ 1.
4
Let Ω ⊆ Rn be a given set, and A be a subset of Ω with A ∈ F. We are interested in the
probability that a ‘randomly chosen point’ in Ω which falls in A. We define
measure(A)
P(A) = .
measure(Ω)
For example, if a point is chosen at random from the interval (a, b), the probability that it lies in
d−c
the interval (c, d), a ≤ c < d ≤ b is b−a . Moreover, the probability that the randomly selected
point lies in any interval of length (d − c) is the same.
Example 1.8. A point is picked ‘at random’ from a unit square. Find the probability that the
point falls in the set of all points which are at most a unit distance from the origin.
Solution: Here Ω = {(x, y) ∈ R2 : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1} and A = {(x, y) ∈ Ω : x2 + y 2 ≤ 1}.
Note that
π12
Area(Ω) = 1, Area(A) = .
4
Thus, the probabilty that the point falls in A is given by
measure(A) π
P(A) = = .
measure(Ω) 4
1.1. Conditional Probability & its properties. Suppose we know that an event B has
happened. Then using this information, we want to measure the chances of happening another
event A i.e., to find P(A|B). Given that the event B occurred, for if it is the case that A occurs
if and only iff A ∩ B occurs. Thus, P(A|B) ∼ P(A ∩ B) i.e., P(A|B) = αP(A ∩ B) for some
constant α = α(B).
Definition 1.7 (Conditional Probability). Let (Ω, F, P) be a probability space. If A, B ∈ F
with P(B) > 0, then the conditional probability of A given B is defined as
P(A ∩ B)
P(A|B) = .
P(B)
Example 1.9. Two fair dice are thrown. Given that the first shows 3, what is the probability
that total exceeds 6?
Solution: Here Ω = {(i, j) : 1 ≤ i, j ≤ 6}, F = P(Ω), P(A) = |A| 36 for any A ⊆ Ω. Let B be the
event that first die shows 3 and A be the event that total exceeds 6, i.e.,
B := {(3, b) : 1 ≤ b ≤ 6}, A := {(a, b) ∈ Ω : a + b ≥ 6}
 PROBABILITY AND STOCHASTIC PROCESS 5

6
Note that P(B) = 36 = 16 .
3 1
A ∩ B = {(3, 4), (3, 5), (3, 6)} =⇒ P(A ∩ B) = = .
36 12
Thus, the required probability is given by
P(A ∩ B) 6 1
P(A|B) = = = .
P(B) 12 2
We now discuss some important properties of conditional probability.
Theorem 1.3. Let (Ω, F, P) be a probability space.
1) Let A ∈ F with P(A) > 0. Then PA := P(·|A) is a probability measure over Ω with
PA (A) = 1.
2) Multiplication rule: If A1 , A2 , . . . , An ∈ F with P(∩n−1
i=1 Ai ) > 0, then
n−1
P(∩ni=1 Ai ) = P(A1 )P(A2 |A1 )P(A3 |A2 ∩ A1 ) . . . P(An | ∩i=1 Ai ).
3. Total probability rule: Let A1 , A2 , . . . be finite or countable partition of Ω i.e.,
Ai ∩ Aj = ∅ ∀i ̸= j, and ∪∞
i=1 Ai = Ω
such that P(Ai ) > 0. Then for any B ∈ F
X∞
P(B) = P(B|Ai )P(Ai ) .
i=1
4. Bayes’ rule: Let A1 , A2 , . . . be finite or countable partition of Ω such that P(Ai ) > 0.
Then for any B ∈ F with P(B) > 0,
P(Ai )P(B|Ai )
P(Ai |B) = P∞ .
j=1 P(B|Aj )P(Aj )
Proof. Proof of 1): Observe that 0 ≤ PA (B) ≤ 1 for any B ∈ F and PA (Ω) = 1. Moreover,
PA (A) = 1. Let {Bi } are disjoint events. Then Ci := Bi ∩ A are disjoint. Hence we have
P∞ ∞ ∞
P A ∩ (∪∞


∞ i=1 Bi ) P(⊔∞i=1 Ci ) P(Ci ) X P(Ci ) X
PA (∪i=1 Bi ) = = = i=1 = = PA (Bi ).
P(A) P(A) P(A) P(A)
i=1 i=1
This show that PA := P(·|A) is a probability measure over Ω.
Proof of 2): Let us take A0 = Ω for convenient. Then, we have
n−1
P(A1 )P(A2 |A1 )P(A3 |A2 ∩ A1 ) . . . P(An | ∩i=1 Ai )
n n r−1

n
Y
r−1
Y P Ar ∩ (∩ Ak )
k=1
Y P(∩rk=1 Ak )
= P(Ar | ∩k=1 Ak ) = r−1 = r−1 = P(∩ni=1 Ai ) .
r=1 r=1
P(∩k=1 Ak ) r=1
P(∩k=1 Ak )
Proof of 3): Observe that, since {Ai } is disjoint, {B ∩ Ai } is also disjoint. Moreover,
B = B ∩ Ω = B ∩ (∪∞ ∞
i=1 Ai ) = ⊔i=1 (B ∩ Ai ).
Thus, by using the property of P, we obtain
∞
X ∞
X
⊔∞


P(B) = P i=1 (B ∩ Ai ) = P(B ∩ Ai ) = P(B|Ai )P(Ai ) .
i=1 i=1
Proof of 4): By using the definition of conditional probability and the total probability rule 2),
we have
P(Ai ∩ B) P(Ai )P(B|Ai ) P(Ai )P(B|Ai )
P(Ai |B) = = = P∞ .
P(B) P(B) j=1 P(B|Aj )P(Aj )
□
6 A. K. MAJEE

Example 1.10. Three teenagers want to get into a R-rated movie. At the box office, ther are
asked to produce their IDs; after the clerk checks them and denies them to enter, he returns the
IDs randomly. Find the probability that none of the teenagers get their own ID.
Solution: Let A be the event that none of the teenagers get their own ID and Ai denotes the
event that i-th teenager gets his own ID. Our required probability is P(A). Note that since there
are three possible cases and only one is favourable, P(Ai ) = 1/3 for i = 1, 2, 3. Note that after
giving the right ID to i-th teenager, probability that j-th teenager will get his own ID is equal to
1/2 as there are two possible ones in which one is favourable. Thus, for i ̸= j, we see that
1 1 1
P(Ai ∩ Aj ) = P(Ai )P(Aj |Aj ) = × = .
3 2 6
In a similar fashion, we get
P(∩3i=1 Ai ) = P(A1 )P(A2 |A1 )P(A3 |(A1 ∩ A2 )) = 1/3 × 1/2 × 1 = 1/6.
Hence , we compute
3
X X
P(A) = P(∩3i=1 A∁i ) =1− P(∪3i=1 Ai ) =1− P(Ai ) + P(Ai ∩ Aj ) − P(∩3i=1 Ai )
i=1 i<j
= 1 − (3 × 1/3) + (3 × 1/6) − 1/6 = 2/6 = 1/3.
Example 1.11. An insurance company believes that people can be divided into two classes: those
who are accident prone and those who are not. During any given year, an accident-prone person
will have an accident with probability 0.4, whereas the corresponding figure for a person who is
not prone to accidents is 0.2. If we assume that 30 percent of the population is accident prone
a) What is the probability that a new policyholder will have an accident within a year of
purchasing a policy?
b) What is the probability that policyholder is accident prone given that he/she has an
accident within a year of purchasing a policy?
c) What is the probability that a new policyholder will have accident in his/her second year
of policy ownership given that the policyholder has had accident in the first year?
Solution: Let A be the event that the policyholder is accident prone and Ai be the event that the
policyholder has had accident in i-th year. We need find P(A1 ), P(A|A1 ) and P(A2 |A1 ). From
the given conditions, we have
P(A1 |A) = 0.4 = P(A2 |A ∩ A1 ), P(A1 |A∁ ) = 0.2 = P(A2 |A∁ ∩ A1 ), P(A) = 0.3, P(A∁ ) = 0.7
By the total probability rule, we obtain
P(A1 ) = P(A1 |A)P(A) + P(A1 |A∁ )P(A∁ ) = (0.4)(0.3) + (0.2)(0.7) = 0.26.
Moreover, by using definition of conditional probability, we get
P(A ∩ A1 ) P(A1 |A)P(A) (0.3)(0.4) 6
P(A|A1 ) = = = = .
P(A1 ) P(A1 ) 0.26 13
Note that PA1 (·) := P(·|A1 ) is a probability measure. Thus, by total probability rule, we have
P(A2 |A1 ) = PA1 (A2 ) = PA1 (A2 |A)PA1 (A) + PA1 (A2 |A∁ )PA1 (A∁ )
= P(A2 |(A ∩ A1 ))P(A|A1 ) + P(A2 |(A∁ ∩ A1 ))P(A∁ |A1 )
= P(A2 |(A ∩ A1 ))P(A|A1 ) + P(A2 |(A∁ ∩ A1 )) 1 − P(A|A1 )
 

6 7
= (0.4) + (0.2) ≈ 0.29.
13 13
 PROBABILITY AND STOCHASTIC PROCESS 7

1.2. Independence of Events and its properties: Sometimes the occurrence of an event B
does not affect the probability of an event A i.e., P(A|B) = P (A). In that case, we say the event
A is ”independent” from event B.
Definition 1.8. We make the following definition regarding various notion of independent of
events:
i) Independent Events: Two events A and B are said to be independent if
P(A ∩ B) = P(A)P(B).
ii) Pairwise Independent Events: A family of events {Ai : i ∈ I} is said to be pairwise
independent if
P(Ai ∩ Aj ) = P(Ai )P(Aj ) ∀ i ̸= j.
iii) Independent family: A family of events {Ai : i ∈ I} is said to be independent /
mutually independent if for every finite subset J ̸= ∅ of I, there holds
Y
P(∩j∈J Aj ) = P(Aj ).
j∈J

Example 1.12. Two fair dice are thrown. Consider the following events: let A be the events
that first die shows 2, B be the event that second die shows 5 and C be the event that total is
equal to 7, i.e.,
A = {(2, b) : 1 ≤ b ≤ 6}, B = {(a, 5) : 1 ≤ a ≤ 6}, C = {(a, b) : a + b = 7, 1 ≤ a, b ≤ 6}
Here Ω = {(i, j) : 1 ≤ i, j ≤ 6}, F = P(Ω), P(D) = |D|
36 for any D ⊆ Ω. It is easy to check that
1 1 1
P(A) = P(B) = P(C) = , P(A ∩ B) = P(B ∩ C) = P(C ∩ A) = , P(A ∩ B ∩ C) = .
6 36 36
Hence
a) The family {A, B, C} is pairwise independent.
1 1
b) The family {A, B, C} is NOT mutually independent as P(A ∩ B ∩ C) = 36 ̸= 216 =
P(A)P(B)P(C).