Conditional Probability

DEFINITION For any two events A and B with P(B)  0, the conditional probability of A
given that B has occurred is defined by
P(A  B)
P(A⏐B)  
P(B)

Example Finding Conditional Probabilities

The table shows the results of a study in which researchers examined
a child’s IQ and the presence of a specific gene in the child.
Find the probability that a child has a high IQ, given that the child has the
gene.
Gene Gene not
present present Total
High IQ 33 19 52
Normal IQ 39 11 50
Total 72 30 102

There are 72 children who have the gene. So, the sample space consists of
these 72 children, as shown at the left. Of these, 33 have a high IQ. So,
33
P1B  A2 = ≈ 0.458.
72
The probability that a child has a high IQ, given that the child has the gene, is
about 0.458.

Example Suppose that of all individuals buying a certain digital camera, 60% include an optional
memory card in their purchase, 40% include an extra battery, and 30% include both a
card and battery.
Consider randomly selecting a buyer and let A  {memory card purchased} and
B  {battery purchased}. Then P(A)  .60, P(B)  .40, and
P(both purchased)  P(A  B)  .30. Given that the selected individual purchased
an extra battery, the probability that an optional card was also purchased is

P(A  B) .30
P(A⏐B)      .75
P(B) .40
That is, of all those purchasing an extra battery, 75% purchased an optional memory
card. Similarly,
P(A  B) .30
P(battery⏐memory card)  P(B⏐A)      .50
P(A) .60

Notice that P(A⏐B)  P(A) and P(B⏐A)  P(B). ■

Example A news magazine publishes three columns entitled “Art” (A), “Books” (B), and
“Cinema” (C). Reading habits of a randomly selected reader with respect to these
columns are

Read regularly A B C AB AC BC ABC

Probability .14 .23 .37 .08 .09 .13 .05

Figure illustrates relevant probabilities.

A B

.02 .03 .07

.05
.04 .08
.20
.51 C

Figure Venn diagram for Example

We thus have

P(A  B) .08
P(A⏐B)      .348
P(B) .23
P(A  (B  C)) .04  .05  .03 .12
P(A⏐B  C)        .255
P(B  C) .47 .47
P(A  (A  B  C))
P(A⏐reads at least one)  P(A⏐A  B  C)  
P(A  B  C)
P(A) .14
     .286
P(A  B  C) .49

and
P((A  B)  C) .04  .05  .08
P(A  B⏐C)      .459 ■
P(C) .37

Two events are independent when the occurrence of one of the events does
not affect the probability of the occurrence of the other event. Two events A
and B are independent when
P1B 0 A2 = P1B2 or when P1A 0 B2 = P1A2.
Events that are not independent are dependent.

Example Classifying Events as Independent or Dependent

Determine whether the events are independent or dependent.
Driving over 85 miles per hour 1A2, and then getting in a car accident 1B2

Solution
Driving over 85 miles per hour increases the chances of getting in an
accident, so these events are dependent.
 The Multiplication Rule for P(A  B)
The definition of conditional probability yields the following result

The Multiplication Rule

P(A  B)  P(A⏐B) ⭈ P(B)

The probability that two events A and B will occur in sequence is

P1A and B2 = P1A2 # P1B 0 A2.
If events A and B are independent, then the rule can be simplified to
P1A and B2 = P1A2 # P1B2. This simplified rule can be extended to any
number of independent events.

Example Four individuals have responded to a request by a blood bank for blood donations. None
of them has donated before, so their blood types are unknown. Suppose only type O
is desired and only one of the four actually has this type. If the potential donors are
selected in random order for typing, what is the probability that at least three individu-
als must be typed to obtain the desired type?

Making the identification B  {first type not O} and A  {second type not
O}, P(B)  4. Given that the first type is not O, two of the three individuals left are
3

not O, so P(A⏐B)  3. The multiplication rule now gives

2

P(at least three individuals are typed)  P(A  B)

 P(A⏐B) ⭈ P(B)
2 3 6
⭈
3 4 12
 .5 ■

Example A chain of video stores sells three different brands of DVD players. Of its DVD
player sales, 50% are brand 1 (the least expensive), 30% are brand 2, and 20% are
brand 3. Each manufacturer offers a 1-year warranty on parts and labor. It is known
that 25% of brand 1’s DVD players require warranty repair work, whereas the cor-
responding percentages for brands 2 and 3 are 20% and 10%, respectively.

1. What is the probability that a randomly selected purchaser has bought a brand 1
DVD player that will need repair while under warranty?
2. What is the probability that a randomly selected purchaser has a DVD player that
will need repair while under warranty?
3. If a customer returns to the store with a DVD player that needs warranty repair work,
what is the probability that it is a brand 1 DVD player A brand 2 DVD player?
A brand 3 DVD player?
 The first stage of the problem involves a customer selecting one of the three
brands of DVD player. Let Ai  {brand i is purchased}, for i  1, 2, and 3. Then
P(A1)  .50, P(A2)  .30, and P(A3)  .20. Once a brand of DVD player is selected,
the second stage involves observing whether the selected DVD player needs war-
ranty repair. With B  {needs repair} and B  {doesn’t need repair}, the given
information implies that P(B⏐A1)  .25, P(B⏐A2)  .20, and P(B⏐A3)  .10.
The tree diagram representing this experimental situation is shown in Figure . The
initial branches correspond to different brands of DVD players; there are two
second-generation branches emanating from the tip of each initial branch, one for
“needs repair” and the other for “doesn’t need repair.” The probability P(Ai) appears
on the ith initial branch, whereas the conditional probabilities P(B⏐Ai) and P(B⬘⏐Ai)
appear on the second-generation branches. To the right of each second-generation
branch corresponding to the occurrence of B, we display the product of probabilities
on the branches leading out to that point. This is simply the multiplication rule in
action. The answer to the question posed in 1 is thus P(A1  B)  P(B⏐A1) ⭈
P(A1)  .125. The answer to question 2 is
P(B)  P[(brand 1 and repair) or (brand 2 and repair) or (brand 3 and repair)]
 P(A1  B)  P(A2  B)  P(A3  B)
 .125  .060  .020  .205

P(B  A1)  P(A1)  P(B  A1)  .125

A 1)  .25
P(B 
ir
Repa
P(B'
A)
1 .
75
.50
) No re
pair
A1
P( d1
an P(B  A2)  P(A2)  P(B  A2)  .060
Br .20
A 2) 
P(B 
P(A2)  .30 ir
Repa
Brand 2 P(B'
 A2 ) 
.80
P( No re
A pair
3) 
Br .20
an P(B  A3)  P(A3)  P(B  A3)  .020
d3 .10
A 3) 
P(B 
ir
Repa
P(B'
 A3 ) 
.90
No re
pair
P(B)  .205

Finally,
P(A1  B) .125
P(A1⏐B)      .61
P(B) .205
P(A2  B) .060
P(A2⏐B)      .29
P(B) .205
and
P(A3⏐B)  1  P(A1⏐B)  P(A2⏐B)  .10
 The initial or prior probability of brand 1 is .50. Once it is known that the
selected DVD player needed repair, the posterior probability of brand 1 increases to
.61. This is because brand 1 DVD players are more likely to need warranty repair
than are the other brands. The posterior probability of brand 3 is P(A3⏐B)  .10,
which is much less than the prior probability P(A3)  .20. ■

Bayes’ Theorem
The computation of a posterior probability P(Aj⏐B) from given prior probabilities
P(Ai) and conditional probabilities P(B⏐Ai) occupies a central position in elementary
probability. The general rule for such computations, which is really just a simple
application of the multiplication rule, goes back to Reverend Thomas Bayes, who
lived in the eighteenth century. To state it we first need another result. Recall that
events A1, . . . , Ak are mutually exclusive if no two have any common outcomes. The
events are exhaustive if one Ai must occur, so that A1  . . .  Ak  S.

Bayes’ Theorem
Suppose that a sample space S is a union of mutually disjoint events B1 , B2 , B3 , . . . ,
Bn , suppose A is an event in S, and suppose A and all the B1 have nonzero probabil-
ities. If k is an integer with 1 ≤ k ≤ n, then
P(A | Bk )P(Bk )
P(Bk | A) =
P(A | B1 )P(B1 ) + P(A | B2 )P(B2 ) + · · · + P(A | Bn )P(Bn )

Bayes’ Theorem
Let A1, A2, . . . , Ak be a collection of k mutually exclusive and exhaustive events
with prior probabilities P(Ai ) (i  1, . . . , k). Then for any other event B for
which P(B)  0, the posterior probability of Aj given that B has occurred is
P(Aj  B) P(B⏐Aj)P(Aj)
P(Aj⏐B)     k j  1, . . . , k
P(B)
冱 P(B⏐Ai) ⭈ P(Ai)
i1
 The Law of Total Probability
Let A1, . . . , Ak be mutually exclusive and exhaustive events. Then for any
other event B,

P(B)  P(B⏐A1)P(A1)  . . .  P(B⏐Ak)P(Ak)

k
 冱 P(B⏐Ai)P(Ai)
i1

Proof Because the Ai s are mutually exclusive and exhaustive, if B occurs it must
be in conjunction with exactly one of the Ai’s. That is, B  (A1  B)  . . . 
(Ak  B), where the events (Ai  B) are mutually exclusive. This “partitioning of B”
is illustrated in Figure. Thus

k k

as desired.
P(B)  冱 P(Ai  B)  冱 P(B⏐Ai)P(Ai)
i1 i1

A1 A3

A4
A2

Partition of B by mutually exclusive and exhaustive Ai’s

Example Incidence of a rare disease. Only 1 in 1000 adults is afflicted with a rare disease for
which a diagnostic test has been developed. The test is such that when an individual
actually has the disease, a positive result will occur 99% of the time, whereas an
individual without the disease will show a positive test result only 2% of the time. If
a randomly selected individual is tested and the result is positive, what is the proba-
bility that the individual has the disease?
To use Bayes’ theorem, let A1
{individual has the disease}, A2
{individ-
ual does not have the disease}, and B
{positive test result}. Then P(A1)
.001,
P(A2)
.999, P(B⏐A1)
.99, and P(B⏐A2)
.02. The tree diagram for this prob-
lem is in Figure:
P(A1  B)
.00099
.99
st
Te
B

.01
.001
se B'

d isea Tes
Has t
A1

.999 P(A2  B)
.01998
A2
.02
Doe
sn't st
hav Te
e di
seas B

e .98
B'

Tes
t
Next to each branch corresponding to a positive test result, the multiplication rule
yields the recorded probabilities. Therefore, P(B)
.00099  .01998
.02097,
from which we have
P(A1  B) .00099
P(A1⏐B)


.047
P(B) .02097
Example Most medical tests occasionally produce incorrect results, called false positives and false
negatives. When a test is designed to determine whether a patient has a certain disease,
a false positive result indicates that a patient has the disease when the patient does not
have it. A false negative result indicates that a patient does not have the disease when the
patient does have it.
When large-scale health screenings are performed for diseases with relatively low
incidence, those who develop the screening procedures have to balance several consid-
erations: the per-person cost of the screening, follow-up costs for further testing of false
positives, and the possibility that people who have the disease will develop unwarranted
confidence in the state of their health.
Consider a medical test that screens for a disease found in 5 people in 1,000. Suppose
that the false positive rate is 3% and the false negative rate is 1%. Then 99% of the time a
person who has the condition tests positive for it, and 97% of the time a person who does
not have the condition tests negative for it. (See exercise 4 at the end of this section.)
a. What is the probability that a randomly chosen person who tests positive for the dis-
ease actually has the disease?
b. What is the probability that a randomly chosen person who tests negative for the dis-
ease does not indeed have the disease?