Probability Measure and
Conditional Probability
Probability Measure
• Probability Measure:
Gives a precise measure of the size/volume of the sets
More precisely a Probability measure 𝑃 is a function over an event space 𝐹
𝑃: 𝐹 → 0,1 such that:
𝑃 𝜙 =0
𝑃 Ω =1
For pair-wise disjoint sets (events) 𝐴 , 𝐴 , 𝐴 , … ∈ 𝐹, 𝑃(𝐴 ⋃𝐴 ⋃𝐴 ⋃ … ) = 𝑃(𝐴 ) +
𝑃(𝐴 ) + 𝑃(𝐴 ) + ⋯
The size of the union of disjoint sets is required to be the sum of their individual sizes, even
for an infinite sequence of disjoint sets
In simple words: the probability measure on a sample space Ω assigns every event 𝐴 ⊆ Ω, a
number in [0, 1], such that 𝑃(Ω) = 1 and 𝑃(𝐴 ∪ 𝐵) = 𝑃(𝐴) + 𝑃(𝐵) for disjoint 𝐴, 𝐵
In simple words: this number ∈ [0, 1] assigned to an event is nothing but the probability/chance
that the event occurs
Also from the above statements one can easily infer that for any event 𝐴 , 𝑃 𝐴 ≥0
These are known as the axioms of Probability
Probability Measure
• Properties of Probability Measure:
Probability of the complement of an event: 𝑃 𝐴̅ = 1 − 𝑃(𝐴)
Probability of the union of two events 𝐴, 𝐵: 𝑃 𝐴⋃𝐵 = 𝑃 𝐴 + 𝑃 𝐵 − 𝑃(𝐴⋂𝐵)
Difference of events: 𝑃 𝐴 − 𝐵 = 𝑃 𝐴 − 𝑃(𝐴⋂𝐵)
• Probability Space:
A triple (Ω, 𝐹, 𝑃) where:
Ω is the sample space (Space of all possible outcomes (elementary events) of a random
experiment)
𝐹 is the event space (Space of all possible events)
𝑃 is the probability measure on (Ω, 𝐹) with 𝑃 Ω = 1
Example:
Tossing a fair coin, what is Ω? What is 𝐹? What is 𝑃?
Roll fair die: 𝑃({1}) = 1/6. All outcomes equally likely. 𝑃 1, 2, 3 =
Toss a drawing pin: 𝑃 𝑝𝑖𝑛 𝑝𝑜𝑖𝑛𝑡𝑖𝑛𝑔 𝑢𝑝𝑤𝑎𝑟𝑑𝑠 ≠ 1/2. All outcomes not equally likely
Example: Dart (Board) Game on Sample Space Ω
Example 2 on Sample Space Ω
• Two persons have a business meeting at a given time. Each person
will arrive at a meeting place with a delay between 0 to 1 hour, with
all pair of delays being equally likely.
The first to arrive will wait for 15 minutes and will leave if the other
has not yet arrived. What is the probability that they meet ?
• If X is the first to arrive, and he arrives after 45min, they are
guaranteed to meet. This probability is, ¼.
• A = P(X arrives within the first 45min) = 45/60 = 3/4.
B = P(Y arrives within some 15min interval) = 15/60 = 1/4.
• P(A and B) = 3/4 * 1/4 = 3/16.
• Hence, overall probability, 3/16 + ¼ = 7/16
Conditional Probability: Introduction
• Suppose for a random experiment 𝜀, we are aware of the correspond-
ing sample space and the probability law
• Also suppose the outcome belongs to some given event 𝐵
• Now, one is required to quantify the likelihood that the outcome also
belongs to some other event 𝐴
In other words, one will have to derive the “New Probability Law” based on
the “available knowledge”
This new probability law will specify the conditional probability of 𝐴 given
(the available knowledge provided through) 𝐵 which is denoted as 𝑃(𝐴|𝐵)
P(E|F)
Probability that
Event E occurs
given that sample
space has been
narrowed down to
the point in F
Source: Harchol-Balter, Mor. Introduction to probability for computing. Cambridge University Press,
2023.
Introduction
• Also given that the outcomes are equally likely, then one possible way
# ∩
to express 𝑃(𝐴|𝐵) could be: 𝑃 𝐴 𝐵 =
#
( ∩ )
• Generalizing the above: 𝑃 𝐴 𝐵 = ; (Here 𝑃 𝐵 > 0)
( )
• What one can say about 𝑃 𝐵 𝐴 ?
• Note: The above formula holds true for countable events
• Conditional Probability constitutes valid probability law, how ?
Example 1
• Six possible outcomes of a fair die roll are equally likely. Now we are
given the information that the outcomes are “odd”, what is the
probability of getting 3 ?
• This implies that:- 𝑃 𝑇ℎ𝑒 𝑜𝑢𝑡𝑐𝑜𝑚𝑒 = 3 𝑇ℎ𝑒 𝑜𝑢𝑡𝑐𝑜𝑚𝑒 𝑖𝑠 𝑜𝑑𝑑) =
Example
• How much do vaccines help? (Exercise 2.3 MOR HARCHOL-BALTER)
The US Surgeon General recently declared that 99.5% of COVID
deaths are among the unvaccinated. Given that I’m vaccinated, what
are my chances of dying from COVID?
You may use any of the facts below:
The fraction of people who are currently vaccinated in the United States is
50%.
The fraction of people who die from COVID in the United States is 0.2%.