Probability and Statistics

Daniel Halbheer
[email protected]
 Session Outline
This session illustrates how probability and
statistics can be used in business settings:
• Predicting outcomes under uncertainty
about the business environment.
• Learning about the business environment
from sample data.
• Making optimal data-driven decisions.

2
 Setting the Stage

Monty Hall problem: Suppose you are on a game show, and you
are given the choice among three doors: Behind one door is a
car; behind the others, goats. You pick a door, say No. 1, and the
host, who knows what's behind the doors, opens another door,
say No. 3, which has a goat. He then says to you, “Do you want to
pick door No. 2?” Is it to your advantage to change your choice?
Source: Wikipedia, link: https://bit.ly/2mXP6vz

3
 Levels of Uncertainty

Levels: Alternative futures (first-order uncertainty), range of

futures (second-order uncertainty), and “true ambiguity.”
4
 Backdrop
Probability: In the study of probability, we
work with a known population—the sample
space of the experiment—and consider the
likelihood of drawing a particular sample.

Statistics: In statistics, we make an inference

about a population by using a sample of data
drawn from that population.

5
What is meant by the probability
of an event? How is it computed?
 Sample Spaces and Events
Definition: A sample space 𝑆 is the space of
all possible outcomes for an experiment. The
elements 𝑠𝑖 of 𝑆 are called sample points. The
number of sample points is denoted by #(𝑆).

Definition: An event 𝐸 for an experiment is a

subset of the sample space. If the outcome of
the experiment is a sample point in 𝐸, then
the event 𝐸 is said to occur.
7
 Coin Tossing
Experiment: A coin is tossed three times
and the result for each toss is observed.

Exercise 1: Let 𝐻 or 𝑇 be the result for each

coin and determine a sample space 𝑆.
Exercise 2: Determine the following events:
▪ 𝐸 = one head and two tails
▪ 𝐹 = head on the first toss

8
 Venn Diagrams
𝑆
𝐸 𝑇𝐻𝐻
𝐹
𝑇𝑇𝐻
𝐻𝐻𝐻
𝑇𝐻𝑇 𝐻𝑇𝑇
𝐻𝐻𝑇
𝐻𝑇𝐻
𝑇𝑇𝑇

Note: If the outcome 𝐻𝑇𝑇 occurs, then both 𝐸

and 𝐹 must occur, and conversely.
9
 Probability
Definition: If all outcomes of an experiment
are equally likely, then the probability of the
event 𝐸 = 𝑠1 , … , 𝑠𝑘 is given by
#(𝐸)
𝑃 𝐸 = 𝑃 𝑠1 + ⋯ + 𝑃 𝑠𝑘 = .
#(𝑆)
Note: 𝑃 𝐸 can be interpreted as the relative
frequency of 𝐸 occurring in the long run.

10
Empirical Probability

11
 Probability Rules
1. For any event 𝐸, 0 ≤ 𝑃(𝐸) ≤ 1.
2. For the set of all outomes 𝑆, 𝑃 𝑆 = 1.
3. If 𝐸 and 𝐹 are events,
𝑃 𝐸∪𝐹 =𝑃 𝐸 +𝑃 𝐹 −𝑃 𝐸∩𝐹 .
4. For any event 𝐸, 𝑃 𝐸 𝑐 = 1 − 𝑃 𝐸 .
Note: 𝐸 𝑐 is the complement of 𝐸—the set of
sample points in 𝑆 that are not in 𝐸.

12
 Conditional Probability
Definition: The conditional probability of
an event 𝐸 given that event 𝐹 has occurred
is given by
𝑃 𝐸∩𝐹
𝑃 𝐸|𝐹 = .
𝑃 𝐹

Note: 𝐸 and 𝐹 are independent if 𝑃 𝐸|𝐹 =

𝑃 𝐸 , that is, if 𝑃 𝐸 ∩ 𝐹 = 𝑃 𝐸 𝑃 𝐹 .

13
 Defective Computer Chips
A firm uses one computer chip to build one
unit of a product. The chips are purchased
from suppliers 𝐴, 𝐵, and 𝐶 and are randomly
picked for assembling a unit. 20% come from
𝐴, 30% come from 𝐵, and the remainder
come from 𝐶. The breakdown probabilities
are 0.03, 0.04, and 0.01, respectively. If a
product is chosen at random, what is the
probability that the chip in it is defective?
14
 Bayes’ Rule
The conditional probability of an event 𝐸
given that event 𝐹 has occurred can be
expressed as
𝑃 𝐹|𝐸 𝑃(𝐸)
𝑃 𝐸|𝐹 = 𝑐 𝑐
.
𝑃 𝐹|𝐸 𝑃 𝐸 + 𝑃 𝐹|𝐸 𝑃(𝐸 )

Note: This result has many applications in

decision making, statistics, and recently AI.

15
 Bayes’ Rule in AI

Read more on the application of Bayes’ rule in “An executive’s guide to AI” published by
McKinsey. Link: https://bit.ly/3AUQs4F.
16
 Deriving Bayes’ Rule
𝑆 Recall: The conditional
𝐸 probability of 𝐸 given
𝐹
that 𝐹 has occurred is
𝐸∩𝐹
𝐸𝑐 ∩ 𝐹 given by
𝑃 𝐸∩𝐹
𝑃 𝐸|𝐹 = .
𝑃 𝐹

Using this definition,

𝑃 𝐹|𝐸 𝑃(𝐸) 𝑃 𝐹|𝐸 𝑃(𝐸)
𝑃 𝐸|𝐹 = = 𝑐 𝑐
.
𝑃(𝐹) 𝑃 𝐹|𝐸 𝑃 𝐸 + 𝑃 𝐹|𝐸 𝑃(𝐸 )
After test result: Before test result:
P(HIV+ | Tested positive) P(Tested positive | HIV+)
17
 Application to HIV Testing
An HIV test gives a positive result with
probability 98% when a patient is indeed
affected by HIV, while it gives a negative
result with probability 99% when a patient
is not affected by HIV. If a person drawn at
random from a population in which 0,1% of
individuals are affected by HIV and she/he is
found positive, what is the probability that
she/he is actually affected by HIV?
18
COVID-19 Tests

19
 Monty Hall Revisited
Suppose you pick door No. 1 and Monty
shows you a goat behind door No. 3.
Define the events 𝐴 = {Car behind door 2}
and 𝐵 = Monty opens door 3 . Then,
1
𝑃 𝐵|𝐴 𝑃(𝐴) 1∗ 2
𝑃 𝐴|𝐵 = = 3 = .
𝑃 𝐵 1 1 1 2 3
1∗ + ∗ +0
3 2 2 3
𝑃(𝐵|𝐴𝑐 ) 𝑃(𝐴𝑐 )
20
What is a random variable? What
is a distribution function?
 Random Variables
If a coin is tossed three times, the number of
heads that occur is captured by the events
0 1 2 3.
Let 𝑋 represent the number of heads and let
𝑋 = 𝑥 be a realization, where 𝑥 = 0,1,2,3.
Note: The realization of 𝑋 is determined by
chance—thus making it a random variable.

22
 Expected Value
Let 𝑋 be a discrete random variable whose
number of possible values is 𝑥1 , … , 𝑥𝑁 .
Definition: The expectation of 𝑋 is
𝑁

𝜇 = 𝐸 𝑋 = ෍ 𝑥𝑖 𝑃(𝑋 = 𝑥𝑖 ) ,
𝑖=1

where 𝑃(𝑋 = 𝑥𝑖 ) is the probability that 𝑋

takes value 𝑥𝑖 .
23
 Variance
𝐸 𝑋 is the central tendency of 𝑋 but does
not indicate the dispersion from the mean.
Definition: The variance of 𝑋 is
𝑁

σ2 = Var 𝑋 = ෍ 𝑥𝑖 − 𝐸(𝑋) 2 𝑃 𝑋 = 𝑥𝑖 .
𝑖=1

The standard deviation of 𝑋 is 𝜎 = Var 𝑋 .

24
 Bernoulli Distribution
Definition: A random variable 𝑋 has a
Bernoulli distribution with parameter
𝑝 ∈ [0,1] if
𝑃 𝑋 =1 =𝑝 =1−𝑃 𝑋 =0 .
Note: A Bernoulli distributed 𝑋 is the
indicator function of the event 𝑋 = 1 .
Exercise: Calculate 𝐸 𝑋 and Var 𝑋 .
25
 Coin Tossing
Experiment: A coin is tossed three times. Let
𝑋 be the number of heads that occur.

Exercise 1: Graph the probability function

𝑃 𝑋 = 𝑥𝑖 .
Exercise 2: Show that 𝐸 𝑋 = 1.5.
Exercise 3: Show that Var 𝑋 = 0.75. What
is the standard deviation of 𝑋? Explain.
26
 Continuous Random Variables
A random variable 𝑋 is continuous if it can
assume any value in an interval.
Definition: The probability that 𝑋 assumes a
value between 𝑎 and 𝑏 is
𝑏
𝑃 𝑎 ≤ 𝑋 ≤ 𝑏 = න 𝑓 𝑥 𝑑𝑥,
𝑎

where 𝑓 𝑥 is the density function of 𝑋.

27
 Distribution Functions
Definition: The cumulative distribution
function (CDF) of 𝑋 is obtained as
𝑥
𝐹 𝑥 = 𝑃 𝑋 ≤ 𝑥 = න 𝑓 𝑡 𝑑𝑡.
−∞

Note: The CDF gives the probability that 𝑋

takes a value less than or equal to 𝑥.
Useful fact: 𝑃 𝑋 ≥ 𝑥 = 1 − 𝐹 𝑥 .

28
 Mean and Variance
Definition: Let 𝑋 be a continuous random
variable. The expectation of 𝑋 is
∞
𝜇 = 𝐸 𝑋 = න 𝑥𝑓 𝑥 𝑑𝑥
−∞

and the variance of 𝑋 is

∞
2 2
σ = Var 𝑋 = න 𝑥−𝐸 𝑋 𝑓 𝑥 𝑑𝑥.
−∞

29
 Uniform Distribution

Note: Probabilities are areas under the density

function and above the 𝑥-axis.
30
 Mean and Variance
Exercise: Let 𝑋 be uniformly distributed on
the interval 0, 𝑏 , where 𝑏 > 0.
Show that
𝑏 𝑏2
𝐸 𝑋 = and Var 𝑋 = .
2 12
How does the parameter 𝑏 affect the mean
and the variance? Explain.

31
 Normal Distribution

Note: The normal distribution is symmetric

around 𝜇 and more spread out for larger 𝜎.
32
 Normal Distribution
Definition: A continuous random variable 𝑋
has a normal distribution if its probability
density function is given by
1 1 𝑥−𝜇 2
−
𝑓 𝑥 = 𝑒 2 𝜎 ,
𝜎 2𝜋
where the parameters 𝜇 and 𝜎 are the mean
and the standard deviation of 𝑋, respectively.

33
 The CDF of 2
𝑁(𝜇, 𝜎 )
The CDF of the normal distribution must be
computed numerically.
Excel: NORM.DIST function

34
 NORM.DIST Function
Exercise: Verify that the probabilities in the
table below are correct.

35
Some Facts

36
Why is the normal distribution so
popular? Who uses it?
 The Hype about 2
𝑁(𝜇, 𝜎 )
The normal distribution is perhaps the most
important of all probability distributions.
Business applications: Quality control, risk
management, stock market returns, employee
performance, measurement error…
Statistics: Sample means tend to be normally
distributed—no matter what the distribution
of the original population is.
38
 Central Limit Theorem

Fact: For 𝑛 ≥ 30, the sample average σ𝑖 𝑥𝑖 /𝑛

has approximately a normal distribution with
mean 𝑝 and variance 𝑝(1 − 𝑝)/𝑛.
39
How can data be used to make
optimal business decisions?
 Data Analytics
Setup: A data analyst collected information
about demand. Extensive testing revealed
that the data is well described by a normal
distribution.
Common normality test: Shapiro-Wilk test.
Business problem: Determine the capacity
such that the probability of a stockout is less
than 2% based on the sample information.
41
 Estimating Parameters
Data analysis: The sample average is
σ𝑖 𝑥𝑖
𝑥ҧ =
𝑛
and the standard deviation is estimated by

σ𝑖(𝑥𝑖 − 𝑥)ҧ 2
𝑠= .
𝑛−1
In Excel: AVERAGE and STDEV, respectively.
42
 Optimal Decisions
Use the data set “stockouts.xlsx” to estimate
the demand parameters and determine the
constrained optimal capacity.
Hint: Use Excel Solver
to find the constrained
optimal capacity.

43
Numerical Optimization

44
 Data-Driven Optimal Capacity

Managerial intuition: Why would a firm try

to limit the likelihood of stockouts? Discuss.
45
 Conclusion
This session illustrated how probability and
statistics can be used in business settings:
• Predicting outcomes under uncertainty
about the business environment.
• Learning about the business environment
from sample data.
• Making optimal data-driven decisions.

46
 Epilogue
In my view, there are two key success factors
in data-rich business environments:
1. Being able to structure links between the
environment, decisions, and outcomes.
2. Being able to make optimal data-driven
decisions.
Next steps: Data Science Camp, core classes.

47
 Feedback
I am interested in learning from you about
your impressions about the class.
Please approach me:
1. In person
2. Via e-mail ([email protected])

Note: Your feedback helps me to improve my

teaching—and to grow as a human being.
48