Chapter 4
Chapter 4
Chapter 4
Fundamentals
of Probability
Theory
Unconditional
Probability
Conditional
Probability
and the
General
Probability Judgment
”And” Rule
Total
Probability
and the
General ”Or”
Rule Manoogian College of Business and Economics, AUA
Bayes’s Rule
Bayesian
Updating
Discussion
1 / 20
Decision-Making under Uncertainty
Introduction
Fundamentals
of Probability 1 Introduction
Theory
Unconditional
Probability 2 Fundamentals of Probability Theory
Conditional
Probability 3 Unconditional Probability
and the
General
”And” Rule 4 Conditional Probability and the General ”And” Rule
Total
Probability
and the 5 Total Probability and the General ”Or” Rule
General ”Or”
Rule
Bayes’s Rule
6 Bayes’s Rule
Bayesian
Updating 7 Bayesian Updating
Discussion
8 Discussion
2 / 20
Introduction
Introduction
Fundamentals
of Probability
Theory
Unconditional
• Although the theory of choice under certainty is helpful for
Probability a range of purposes, most real-life decisions are not
Conditional
Probability
choices under certainty.
and the
General • In this module, we explore the theory of probability that
”And” Rule
many consider as a correct normative theory of
Total
Probability probabilistic judgment.
and the
General ”Or” • Like the theory of rational choice under certainty,
Rule
Bayes’s Rule
probability theory is axiomatic, so we start learning about
Bayesian decision-making under uncertainty from a set of axioms in
Updating
probability.
Discussion
3 / 20
Important Definitions
Introduction
Total
Definition of ”outcome”
Probability
and the
An outcome is a subset of the outcome space.
General ”Or”
Rule
Introduction
Fundamentals
of Probability
Definition of ”equiprobable”
Theory
Outcomes are equiprobable if they occur with the same
Unconditional
Probability probability.
Conditional
Probability
and the Definition of ”mutual exclusivity”
General
”And” Rule Outcomes are mutually exclusive if no more than one of them
Total
Probability
can occur at any one time.
and the
General ”Or”
Rule
Definition of ”independence”
Bayes’s Rule
Bayesian
Outcomes are independent if the occurrence of the one does
Updating not affect the probability of the other one occurring.
Discussion
Note that when two outcomes are mutually exclusive they are
not independent.
5 / 20
Rules of (Unconditional) Probability
Introduction
Fundamentals
of Probability
Theory
Conditional
If all individual outcomes in an outcome space
Probability
and the
{A1 , A2 , · · · , An } are equiprobable, then the probability of any
General
”And” Rule
one individual outcome Ai is 1/n, that is: P r(Ai ) = 1/n.
Total
Probability
and the The ”Or” Rule
General ”Or”
Rule If outcomes A and B are mutually exclusive, then the
Bayes’s Rule probability of A or B equals the probability of A plus the
Bayesian
Updating
probability of B, that is: P r(A ∨ B) = P r(A) + P r(B).
Discussion
6 / 20
Examples
Introduction
Fundamentals
of Probability
Theory
Unconditional
Probability
Conditional
• A roulette wheel has spaces numbered from 1–36
Probability
and the
alternating red and black plus two additional spaces, 0 and
General 00, both colored green. Thus, there are 18 red spaces, 18
”And” Rule
black spaces, and 2 green spaces. What is:
Total
Probability – The probability of getting black?
and the
General ”Or” – The probability of getting black or green?
Rule – The probability of spinning an odd number, 0, or 00?
Bayes’s Rule
Bayesian
Updating
Discussion
7 / 20
Rules of (Unconditional) Probability, cont.
Introduction
Fundamentals
of Probability
Theory
The ”Everything” Rule
Unconditional The probability of the entire outcome space equals 1.
Probability
Conditional
Probability The ”Not” Rule
and the
General
”And” Rule
The probability that some outcome A will not occur equals 1
Total
minus the probability that it does, that is:
Probability
and the
P r(¬A) = 1 − P r(A).
General ”Or”
Rule
Discussion
B equals the probability of A times the probability of B, that
is:P r(A&B) = P r(A) × P r(B).
8 / 20
More Examples
Introduction
Fundamentals
of Probability
Theory
Unconditional
Probability
Conditional
Probability • Consider the roulette wheel again. What is:
and the
General – The probability of not getting a number between 1 and 10?
”And” Rule
– The probability of getting red, black, or green?
Total
Probability – The probability of getting black on the first spin and green
and the
General ”Or”
on the second spin?
Rule
Bayes’s Rule
Bayesian
Updating
Discussion
9 / 20
Conditional Probability
Introduction
Fundamentals
of Probability
Theory
10 / 20
Conditional Probability, cont.
Introduction
Fundamentals
• Example: Suppose that you draw one card from a well-shuffled deck, and
of Probability that you are interested in the probability of drawing the ace of spades given
Theory
that you draw an ace.
Unconditional
Probability • Answer: Given that you just drew an ace, there are four possibilities: the
Conditional ace of spades, ace of clubs, ace of hearts, or ace of diamonds. Because only
Probability
and the one of the four is the ace of spades, and because all four outcomes are
General
”And” Rule equally likely, this probability is 1/4.
Total • You can get the same answer by dividing the probability that you draw the
Probability
and the ace of spades by the probability that you draw an ace: 1/52 divided by
General ”Or” 4/52, which is 1/4.
Rule
Bayes’s Rule
Bayesian
Formal Definition of Conditional Probability
Updating
If A and B are two outcomes, the probability of A
Discussion
conditional of B equals the probability of A and B divided
by the probability of B, that is: P r(A|B) = P Pr(A&B)
r(B)
.
11 / 20
The General ”And” Rule and Independence
Conditions
Introduction
Total
rule because it doesn’t require A and B to be independent.
Probability
and the
• The following equivalence holds:
General ”Or”
Rule
P r(A|B) × P r(B) = P r(B|A) × P r(A)
Bayes’s Rule
Bayesian
• Also, the following independence conditions are logically
Updating
equivalent: A and B are independent if
Discussion
i. P r(A|B) = P r(A)
ii. P r(B|A) = P r(B)
iii. P r(A&B) = P r(A) × P r(B)
12 / 20
Total Probability
Introduction
Fundamentals
of Probability
Theory
Unconditional
Probability
Conditional
Probability
and the
The Rule of Total Probability
General
”And” Rule P r(D) = P r(D&B) + P r(D&¬B)
Total or
Probability
and the P r(D) = P r(D|B) × P (B) + P r(D|¬B) × P r(¬B)
General ”Or”
Rule
Bayes’s Rule
Bayesian
Updating
Discussion
13 / 20
The General ”Or” Rule
Introduction
Conditional
P r(A ∨ B) = P r(A) + P r(B) − P r(A&B).
Probability
and the
General
”And” Rule
• This rule is a generalization of the previously stated ”or”
Total rule because it doesn’t require A and B to be mutually
Probability
and the
exclusive.
General ”Or”
Rule • Example: Suppose you roll two dice. What is the
Bayes’s Rule probability that the first roll results in 2 or the second roll
Bayesian results in 2 (That is 2 comes up at least once)?
Updating
Discussion
• Property: P r(A ∨ B) = 1 − P r(¬A & ¬B)
• Example contd: Answer the question in the above
example using the property above.
14 / 20
Bayes’s Rule
Introduction
Fundamentals
of Probability
Theory
Unconditional
Probability Bayes’s Rule
Conditional P r(D&B)
Probability P r(B|D) = P r(D)
and the
General
”And” Rule or
Total P r(D|B)×P r(B)
Probability P r(B|D) = P r(D)
and the
General ”Or”
Rule
or
P r(D|B)×P r(B)
Bayes’s Rule P r(B|D) = P r(D|B)×P (B)+P r(D|¬B)×P r(¬B)
Bayesian
Updating
Discussion
15 / 20
Example
Introduction
Fundamentals
of Probability
Theory
Unconditional
Probability
Conditional
Probability • Ten percent of a company’s employees are smokers. For
and the
General each nonsmoker, the chance of taking a sick day in the
”And” Rule
next year is 0.01. For each smoker, the chance of taking a
Total
Probability sick day is 0.05. Given that a worker is sick, what is the
and the
General ”Or” chance he was a smoker?
Rule
Bayes’s Rule
Bayesian
Updating
Discussion
16 / 20
Bayesian Updating
Introduction
Fundamentals
• We update beliefs in light of new evidence all the time.
of Probability
Theory
• Bayes’s rule is often interpreted as describing how we
Unconditional should update our beliefs in light of new evidence.
Probability
• Suppose you have a belief or hypothesis. The question is
Conditional
Probability how you should change your belief – that is, the probability
and the
General that you assign to the possibility that the hypothesis is
”And” Rule
true – in light of the fact that the evidence obtains.
Total
Probability • Let H stand for the hypothesis and E for the evidence.
and the
General ”Or” • The probability of H,P r(H), is called the prior
Rule
probability: it is the probability that H is true before you
Bayes’s Rule
learn whether E is true.
Bayesian
Updating • The probability of H given E, P r(H|E), is called the
Discussion posterior probability: it is the probability that H obtains
given that the evidence E is true. The question is what
the posterior probability should be.
17 / 20
Bayesian Updating
Introduction
Fundamentals
of Probability
Theory
• The last question on the previous slide is answered by a
Unconditional
Probability simple application of the Bayes’s rule:
Conditional
Probability
and the P r(E|H) × P r(H)
General P (H|E) =
”And” Rule P r(E|H) × P r(H) + P r(E|¬H) × P r(¬H)
Total
Probability • Thus Bayes’s rule tells you that the probability you assign
and the
General ”Or” to H being true should go from P r(H), prior probability
Rule
Bayes’s Rule
to P r(H|E), posterior probability.
Bayesian • Bayesian updating is the process of updating beliefs in
Updating
accordance with Bayes’s Rule.
Discussion
18 / 20
Bayesian Updating, an Example
Introduction
Fundamentals
of Probability
Theory
Unconditional
Probability
• Wes is trying to determine whether or not a coin has two
Conditional
Probability
and the
heads. He cannot see the coin, he only knows the outcome
General
”And” Rule
after it is flipped. He believes with 20% confidence that it
Total
has two heads. The first flip is heads. What is his updated
Probability
and the
probability? The second flip is also head. What is his
General ”Or”
Rule
updated probability? The third flip is tails. What is
Bayes’s Rule
updated probability?
Bayesian
Updating
Discussion
19 / 20
Discussion
Introduction
• In this chapter we have explored the theory of probability.
Fundamentals
of Probability • Probability theory is critical to a wide range of applications, among other
Theory
things as the foundations of statistical inference.
Unconditional • It is relevant here because it can be interpreted as a theory of judgment,
Probability
Conditional
that is, as a theory of how to revise beliefs in light of evidence.
Probability • How plausible is this theory?
and the
General • Again, we must separate the descriptive question from the normative
”And” Rule
question. Do people in fact update their beliefs in accordance with Bayes’s
Total
Probability
rule? Should they?
and the • The axioms might seem weak and uncontroversial, both from a descriptive
General ”Or”
Rule and from a normative standpoint.
Bayes’s Rule • Yet the resulting theory is anything but weak. As in the case of the theory
Bayesian of choice under certainty, we have built a remarkably powerful theory on
Updating
the basis of a fairly modest number of seemingly weak axioms.
Discussion • The theory is not intended to describe the actual cognitive processes you go
through when updating your beliefs.
• But it does specify exactly how your posterior probability must relate to
your prior, given various conditional and unconditional probabilities. 20 / 20