UIC
BUSINESS
Basic Probability Concepts
(These slides are meant to cover probability concepts at the level required in this course.
For those interested in learning more, you may want to take IDS 570. I am happy to
point you to additional references as well)
IDS 532
Instructor: Selva Nadarajah
Uncertainty and Random Variables
UIC
BUSINESS
We are interested in an event where the outcome is uncertain
The outcome of the stochastic event is associated with a numeric value
Therefore, the stochastic event is called a random variable and each of its outcomes
has both a value and a probability
Examples:
1. Demand for a product:
Outcomes: Demand equals a value in set {0 units, 20 units,., 100 units}
Values: Number of units of demand corresponding to the outcome
2. Reward from good or bad decisions
Outcomes: good or bad
Values: $100 if the outcome is good and -$40 if out come is bad
3. Time before next patient arrival
Outcome: Time equals any number between 1 and 60 mins
Values: Time corresponding to outcome
Selva Nadarajah
Discrete Vs. Continuous Random Variables
UIC
BUSINESS
Random variables corresponding to a stochastic event can either take discrete values
or continuous values
Demand for a product is a discrete random variable
Time before next patient arrival is a continuous random variable
For analysis, it may be convenient to model a stochastic event with discrete outcomes
using continuous random variables
Selva Nadarajah
UIC
BUSINESS
What is a Random Model?
A random model specifies the possible outcomes for a random variable (e.g. demand)
and the probability of these outcomes.
Traditional distributions from statistics can be used as random models:
e.g., the normal, gamma, Poisson distributions
GAMMA
POISSON
Probability
Probability
NORMAL
Selva Nadarajah
50
100
Demand
150
200
9 10
Demand
4
UIC
BUSINESS
Density Function?
We will use ( )to
denote density
functions
Probability
The density function tells you the probability of a particular outcome occurring.
Density functions can take all kinds of shapes (they dont always have to look like a
bell curve)
For discrete random variables the density function is usually called a probability mass
function e.g. the probability demand will be exactly 5 units.
0
Selva Nadarajah
50
100
Demand
150
200
5
UIC
BUSINESS
Distribution Function?
The distribution function tells you the probability the outcome will be a particular
value or smaller.
e.g. the probability demand will be 5 or fewer units.
Distribution functions always start low and increase towards 1.0
1.00
0.80
We will use F( )to
denote density
functions
0.60
Probability
We will most often
work with distribution
functions.
0.40
0.20
0.00
0
Selva Nadarajah
50
100
Demand
150
200
6
Expected Value
UIC
BUSINESS
It is the probability weighted average of the outcomes of a random variable
Example:
Suppose a random variable models the reward from the outcome of a decision being good or bad.
A good decision results in a reward of $100, where as a bad outcome results in a loss of $40.
The good and bad outcomes occur with probabilities 0.25 and 0.75, respectively
What is the expected value of the decision?
$100 good decision + $40 bad decision
= $100 0.25 + 40 0.75 = $5
In short, we computed the expected value by multiplying the probability of an outcome
with the value of each outcome and summing all these calculations
Selva Nadarajah
General expressions for expected value
UIC
BUSINESS
The notation E(D) is typically used to denote the expectation of a random variable D
Discrete random variables
Assume the random variable takes values , , , with probabilities
, , , ( )
= ( )
Continuous random variables
Assume the random variable takes values from to + and has the probability
density function (
)
=
When moving from discrete to continuous random variables we have replaced the sum by
an integral in the definition of the expectation
Selva Nadarajah
Normal Distribution (Continuous)
UIC
BUSINESS
All normal distributions are characterized by two parameters: mean () and standard
deviation ()
Mean is the expected value of the random variable (also loosely referred to as the
average)
Standard deviation measures the dispersion from the mean
All normal distributions are related to the standard normal that has zero mean and
standard deviation equal to 1.
Selva Nadarajah
Computing Probabilities Using Normal Distributions
UIC
BUSINESS
Let be some quantity, and (, ) the parameters of the normal distribution used to
model demand.
Required: Prob demand is or lower
Prob demand is or lower = Prob(the outcome of a standard normal is z or lower)
Referred to as
the z-statistic
z=
or = +
Compute the desired probability using Excel:
Prob(the outcome of a standard normal is z or lower)=NORMSDIST()
Selva Nadarajah
10
UIC
BUSINESS
Poisson Distribution (Discrete)
0.60
0.20
5
0.18
Density function probability
Six Poisson density
functions with means
0.625, 1.25, 2.5,
5,10 and 20
Density function probability
0.50
0.625
0.40
1.25
0.30
0.20
2.5
0.10
0.16
0.14
10
0.12
0.10
20
0.08
0.06
0.04
0.02
0.00
0.00
0
6
Q
10
10
20
30
Defined only by its mean (standard deviation = square root(mean))
Does not always have a bell shape, especially for low demand.
Discrete distribution function: only non-negative integers
Good for modeling demands with low means (e.g., less than 20)
Selva Nadarajah
11
Computing Probabilities Using Poisson Distributions
UIC
BUSINESS
Let be some nonnegative quantity, and be the mean of the Poisson distribution
used to model demand.
Required: Prob demand is or lower
Compute the desired probability using Excel:
Prob demand is or lower = POISSON(, , 1)
Selva Nadarajah
12
UIC
BUSINESS
Expected Loss
Loss for random variable D is define as follows:
D
=
0
, if D
, if D <
= max{0, D }
Expected loss (EL) is then max 0, (see slide 8)
Relevance to operations? When the random variable denotes demand and is the
number of units of an item in stock, expected loss is the same as expected lost sales
Selva Nadarajah
13
Computing Expected Loss
UIC
BUSINESS
For normally distributed demand with mean and standard deviation , we have
EL = , where =
; = (1 )
Now lets assume we have poisson distributed demand, D, taking values , , ,
with probabilities , , , ( ), respectively. To compute expected loss, we
use the definition on slide 8:
= max 0, ( )
Selva Nadarajah
14
Independence and Correlation
UIC
BUSINESS
Two random variables are independent if the outcome of one of the random variables
has no effect on the outcome of the other
Example, a high demand outcome for one variable provides no information about whether
the second variable will have a high or low demand outcome
Two random variables are correlated if the outcome of one variable provides
information about the outcome of the other
Positively correlated: If one variable is high, then the other tends to be high, and if one
variable is low then the other also tends to be low
Negatively correlated: If one variable is high, then the other tends to be low, and if one
variable is low then the other tends to be high
Selva Nadarajah
15
Sum of Independent Normal and Poisson Random Variables UIC
BUSINESS
Suppose we have independent normal or Poisson random variables , , ,
The mean and standard deviation of the i-th random variable are and ,
respectively
We are interested in the mean and standard deviation of the sum of these random
variables, that is, = + + +
The mean and standard deviation of are
= + + +
=
Selva Nadarajah
+ + +
16
Sum of Two Correlated Normal Random Variables
UIC
BUSINESS
Suppose we have two dependent normal random variables and with means
and and standard deviations and
The dependence between these two random variables is modeled by a number
between -1 and 1 referred to as the correlation coefficient
We are interested in the mean and the standard deviation of the sum of these two
random variables, that is, = +
= +
=
Selva Nadarajah
+ 2 +
17
