Probability in Decision Making

(Analysis of nominal/categorical data)

 Random Process (experiment)
A process leading to outcomes which are not
known in advance.
•A coin is tossed
•A consumer is asked which of two products
he or she prefers
•The daily change in an index of stock
market prices is observed
•An employee will leave an organization
within one year from joining

17-12-2020 EPGP13
 Sample Space (Ω)
Collection of all possible outcomes
An event is an outcome from the sample space.
e.g.: All six faces of a die:

e.g.: All 52 cards

a deck of
bridge cards

Mode of payment {cash, debit card, credit card.}

17-12-2020 EPGP13
 Sample Space

Consider the sample space to describe a randomly chosen Bank

employee by:
2 genders
21 job classifications
6 branch bases
4 education levels

There are: 2 x 21 x 6 x 4 = 1008 possible outcomes

17-12-2020 EPGP13
 What is Probability?
1. Numerical measure of
1 Sure
likelihood that the event
will occur
2. Lies between 0 & 1

.5

0 Impossible

17-12-2020 EPGP13
 Statistical Definition of Probability
P(an event occurs)
= no of data (observations) satisfying the
event / total no of observations (data)

Note: this definition uses relative frequency

approach. Total no of observations has to be very
large.

17-12-2020 EPGP13
 Example
A local Dairy Bar is located in a busy mall in Kolkata. It
sells ice cream and frozen yogurt products. One of
the difficulties in this business is knowing how much
of a given product to prepare for the day. The owner
is interested in determining the probability that a
customer will select yogurt over ice cream. She has
maintained records of customer purchases for the
past three weeks.

17-12-2020 EPGP13
The probability can be assessed using relative frequency
Step 1 Define the experiment.
A randomly chosen customer will select between ice cream and
yogurt.
Step 2 Define the events of interest.
The manager is interested in the event E customer selects yogurt.
Step 3 Determine the total number of occurrences.
Suppose, she has observed 2,250 sales of ice cream and yogurt in
the past three weeks. Thus, N = 2,250.
Step 4 For the event of interest, determine the number of
occurrences.
In the past three weeks, 1,570 sales were for yogurt.
Stp5 Determine the probability assessment.
Thus, based on past history, the chance that a customer will
purchase yogurt is just under 0.70.

17-12-2020 EPGP13
 Example 4.4
In a recent quarterly report, Crystal Cable reported that it had
12.4 million cable television customers and 27.4 million cable
passings (ie number of households that the cable company’s
could reach). Consider randomly selecting one of Crystal’s cable
passings. That is, consider selecting one cable passing by giving
each and every cable passing the same chance of being selected.
Let A be the event that the randomly selected cable passing has
Crystal’s cable television service. Then, because the sample
space of this experiment consists of 27.4 million equally likely
sample space outcomes (cable passings) ,it follows that

P(A) = the number of passings that have Crystal’s cable television service
the total number of cable passings
= 0.45

17-12-2020 EPGP13
Why are some mutual fund managers more successful than
others? One possible factor is where the manager earned his or
her MBA. The following table compares mutual fund
performance against the ranking of the school where the fund
manager earned their MBA:
Mutual fund outperforms Mutual fund doesn’t
the market outperform the market
Top 20 MBA program .11 .29
Not top 20 MBA program .06 .54

E.g. This is the probability that a mutual fund

outperforms AND the manager was not in a
top-20 MBA program; it’s a joint
probability.

17-12-2020 EPGP13
Alternatively, we could introduce shorthand notation to
represent the events:
A1 = Fund manager graduated from a top-20 MBA program
A2 = Fund manager did not graduate from a top-20 MBA
program
B1 = Fund outperforms the market
B2 = Fund does not outperform the market

B1 B2
A1 .11 .29
A2 .06 .54

E.g. P(A2 and B1) = P(A2  B1) =.06

= the probability a fund outperforms the market
and the manager isn’t from a top-20 school.
17-12-2020 EPGP13
Marginal Probabilities…
Marginal probabilities are computed by adding across rows and
down columns; that is they are calculated in the margins of the
table:

P(A2) = .06 + .54

B1 B2 P(Ai)
A1 .11 .29 .40
A2 .06 .54 .60
P(Bj) .17 .83 1.00

P(B1) = .11 + .06 BOTH margins must add to 1

“what’s the probability a fund (useful error check)
outperforms the market?”
17-12-2020 EPGP13
Events Has cable Does not have Total
internet cable internet
service, B service, B
Has cable 6.5 5.9 12.4
television
service , A
Does not have 3.3 11.7 15.0
cable television
service, A
Total 9.8 17.6 27.4

17-12-2020 EPGP13
Events Has cable Does not have Total
internet cable internet
service, B service, B
Has cable 0.24 0.21 0.45
television
service , A
Does not have 0.12 0.43 0.55
cable television
service, A
Total 0.36 0.64 1

17-12-2020 EPGP13
 Question
A survey is conducted on 20 respondents to gather information on
customer satisfaction for a product. The data on customer
satisfaction is obtained on a 3 point scale viz. highly satisfied (HS),
satisfied (S), not satisfied (NS) and also on gender- male (M) and
female (F). The data is recorded as shown below:

Responden 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
ts

Gender M F F F F F F M M M M M F F F F M F M M

Satisfaction S S NS S NS NS NS NS HS HS S NS HS S S HS NS NS S S
level

17-12-2020 EPGP13
 Gender
Satisfaction F M Grand Total
HS 13.81% 9.05% 22.86%
NS 18.57% 17.62% 36.19%
S 16.67% 24.29% 40.95%
Grand Total 49.05% 50.95% 100.00%

17-12-2020 EPGP13
 Conditional Probability
1. Event Probability Given that Another Event
Occurred
2. Revise Original Sample Space (becomes
restrictive) to Account for New information
3. Interested to find out number of cases where
both A and B occur out of all the cases where B
occurs
4. P(A | B) = P(A ∩ B) , P(B)>0
P(B)
17-12-2020 EPGP13
 Example
Recall the previous restaurant example,
what is the probability that a ketchup
user also uses mustard?
P(A|B) = P(AB)/P(B)
= .65/.80 = .8125

17-12-2020 EPGP13
Conditional Probability…
What’s the probability that a fund will outperform
the market given that the manager graduated from a
top-20 MBA program?
Recall:
A1 = Fund manager graduated from a top-20 MBA
program
A2 = Fund manager did not graduate from a top-20
MBA program
B1 = Fund outperforms the market
B2 = Fund does not outperform the market
Thus, we want to know “what is P(B1 | A1) ?”
17-12-2020 EPGP13
 Conditional Probability…
We want to calculate P(B1 | A1)

B1 B2 P(Ai)
A1 .11 .29 .40
A2 .06 .54 .60

P(Bj) .17 .83 1.00

Thus, there is a 27.5% chance that that a fund will outperform the market
given that the manager graduated from a top-20 MBA program.
17-12-2020 EPGP13
 Statistical Independence
1. Tests For Independence

P(A | B) = P(A), or P(B | A) = P(B),

or P(A and B) = P(A)P(B)

2. A and B are said to be independent iff

P(A and B) = P(A)P(B)

17-12-2020 EPGP13
A government agency has 6,000 employees. The employees were
asked whether they preferred a four-day work week (10 hours per day),
a five-day work week (8 hours per day), or flexible hours. You are given
information on the employees' responses broken down by sex.

a.
What is the probability that a randomly selected employee is a man
and is in favour of a four-day work week?
b.
What is the probability that a randomly selected employee is female?
c.
A randomly selected employee turns out to be female. Compute the
probability that she is in favorMaleof flexible
Femalehours.Total
d.
What percentageFourof employees
days 300 is in favour
600 of a five-day
900 work week?
Five days 1,200 1,500 2,700
e.
Given that a person is in favor
Flexible 300 of flexible
2,100 time,2,400
what is the probability
that the person isTotal
female? 1,800 4,200 6,000
f.
What percentage of employees is male and in favor of a five-day work
week?

17-12-2020 EPGP13