Continuous random variables

Chrysafis Vogiatzis

Department of Industrial and Enterprise Systems Engineering

University of Illinois at Urbana-Champaign

Lecture 7 b

©Chrysafis Vogiatzis. Do not distribute without permission of the author

1 / 10 Chrysafis Vogiatzis Continuous random variables

Uniform distributed random variables

Not to be confused with its discrete counterpart.

Here X is allowed to take any value in [α, β].
(
1
β−α ,if α ≤ x ≤ β
f (x) =
0, otherwise.

Zx 0,
 if x < α
x−α
F (x) = f (y )dy = β−α , if α ≤ x ≤ β

1, if x > β

−∞

F (x)
f (x)

α β α β

2 / 10 Chrysafis Vogiatzis Continuous random variables

Exponentially distributed random variables

The second continuous random variable distribution we will see is

immensely useful with a vast number of applications.
Typically used to model time to next random event. Examples
include:
time to next machine failure,
time to next accident,
time to next customer arrival,
time to next shark attack.
Closely related to Poisson random variables and the geometric
distribution (which are discrete).
Similarly to Poisson random variables, we need a rate λ > 0.

3 / 10 Chrysafis Vogiatzis Continuous random variables

Exponentially distributed random variables

The second continuous random variable distribution we will see is

immensely useful with a vast number of applications.
Typically used to model time to next random event. Examples
include:
time to next machine failure,
time to next accident,
time to next customer arrival,
time to next shark attack.
Closely related to Poisson random variables and the geometric
distribution (which are discrete).
Similarly to Poisson random variables, we need a rate λ > 0.

3 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution

Given a parameter (called rate) λ > 0, the exponential distribution

has:
λ · e−λx , for x ≥ 0

pdf: f (x) =
0, for x < 0
−λx

1−e , for x ≥ 0
cdf: F (x) =
0, for x < 0

How is it related to the geometric distribution?

In the discrete case, the number of trials to the first success was
modeled as a geometric distribution.
In the continuous case, the time to the first success is modeled as
an exponential distribution.

How is it related to the Poisson distribution?

Time to next event: exponential distribution.
Number of events in some time: Poisson distribution.

4 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution

Given a parameter (called rate) λ > 0, the exponential distribution

has:
λ · e−λx , for x ≥ 0

pdf: f (x) =
0, for x < 0
−λx

1−e , for x ≥ 0
cdf: F (x) =
0, for x < 0

How is it related to the geometric distribution?

In the discrete case, the number of trials to the first success was
modeled as a geometric distribution.
In the continuous case, the time to the first success is modeled as
an exponential distribution.

How is it related to the Poisson distribution?

Time to next event: exponential distribution.
Number of events in some time: Poisson distribution.

4 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution

Given a parameter (called rate) λ > 0, the exponential distribution

has:
λ · e−λx , for x ≥ 0

pdf: f (x) =
0, for x < 0
−λx

1−e , for x ≥ 0
cdf: F (x) =
0, for x < 0

How is it related to the geometric distribution?

In the discrete case, the number of trials to the first success was
modeled as a geometric distribution.
In the continuous case, the time to the first success is modeled as
an exponential distribution.

How is it related to the Poisson distribution?

Time to next event: exponential distribution.
Number of events in some time: Poisson distribution.

4 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution

Given a parameter (called rate) λ > 0, the exponential distribution

has:
λ · e−λx , for x ≥ 0

pdf: f (x) =
0, for x < 0
−λx

1−e , for x ≥ 0
cdf: F (x) =
0, for x < 0

How is it related to the geometric distribution?

In the discrete case, the number of trials to the first success was
modeled as a geometric distribution.
In the continuous case, the time to the first success is modeled as
an exponential distribution.

How is it related to the Poisson distribution?

Time to next event: exponential distribution.
Number of events in some time: Poisson distribution.

4 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution pdf visually

0.8

λ = 0.5
0.6
λ=1
λ = 1.5
0.4 λ=2

0.2

0
0 0.2 0.4 0.6 0.8 1

5 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution pdf visually

1.5
λ = 0.5
λ=1
1 λ = 1.5
λ=2

0.5

0 1 2 3 4 5

5 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution pdf visually

1.5
λ = 0.5
λ=1
1
λ = 1.5
λ=2

0.5

0 1 2 3 4 5

5 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution pdf visually

1.5
λ = 0.5
λ=1
1
λ = 1.5
λ=2
0.5

0 1 2 3 4 5

5 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution pdf visually

1.5
λ = 0.5
λ=1
1
λ = 1.5
λ=2
0.5

0
0 1 2 3 4 5

5 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution cdf visually

0.8

λ = 0.5
0.6
λ=1
λ = 1.5
0.4 λ=2

0.2

0
0 0.2 0.4 0.6 0.8 1

6 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution cdf visually

0.8
λ = 0.5
0.6 λ=1
λ = 1.5
0.4 λ=2

0.2

0 1 2 3 4 5

6 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution cdf visually

0.8
λ = 0.5
0.6 λ=1
λ = 1.5
0.4 λ=2

0.2

0 1 2 3 4 5

6 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution cdf visually

0.8
λ = 0.5
0.6 λ=1
λ = 1.5
0.4 λ=2

0.2

0 1 2 3 4 5

6 / 10 Chrysafis Vogiatzis Continuous random variables

The exponential distribution cdf visually

0.8
λ = 0.5
0.6 λ=1
λ = 1.5
0.4 λ=2

0.2

0 1 2 3 4 5

6 / 10 Chrysafis Vogiatzis Continuous random variables

Expo. and Poisson distributed random variables

We said earlier that these random variables are siblings:

Assume an event that happens in time that is exponentially
distributed with rate λ.
Then, the number of events that have taken place is Poisson
distributed with rate λ.

Example
Historically, an emergency room after hours (10pm–6am) sees 48
requests every 8 hours. The time until the next patient arrives is
exponentially distributed with that rate. What is the probability that the
next patient arrives in the next 10 minutes? What is the probability
there are 5 patients during the next hour?
48 requests
Answer: Let’s use λ = 8 hours = 6 requests per hour.
The time to the next patient arriving is exponentially distributed.
The number of patients is Poisson distributed.

7 / 10 Chrysafis Vogiatzis Continuous random variables

Expo. and Poisson distributed random variables

We said earlier that these random variables are siblings:

Assume an event that happens in time that is exponentially
distributed with rate λ.
Then, the number of events that have taken place is Poisson
distributed with rate λ.

Example
Historically, an emergency room after hours (10pm–6am) sees 48
requests every 8 hours. The time until the next patient arrives is
exponentially distributed with that rate. What is the probability that the
next patient arrives in the next 10 minutes? What is the probability
there are 5 patients during the next hour?
48 requests
Answer: Let’s use λ = 8 hours = 6 requests per hour.
The time to the next patient arriving is exponentially distributed.
The number of patients is Poisson distributed.

7 / 10 Chrysafis Vogiatzis Continuous random variables

Expo. and Poisson distributed random variables

We said earlier that these random variables are siblings:

Assume an event that happens in time that is exponentially
distributed with rate λ.
Then, the number of events that have taken place is Poisson
distributed with rate λ.

Example
Historically, an emergency room after hours (10pm–6am) sees 48
requests every 8 hours. The time until the next patient arrives is
exponentially distributed with that rate. What is the probability that the
next patient arrives in the next 10 minutes? What is the probability
there are 5 patients during the next hour?
48 requests
Answer: Let’s use λ = 8 hours = 6 requests per hour.
The time to the next patient arriving is exponentially distributed.
The number of patients is Poisson distributed.

7 / 10 Chrysafis Vogiatzis Continuous random variables

Expo. and Poisson distributed random variables

We said earlier that these random variables are siblings:

Assume an event that happens in time that is exponentially
distributed with rate λ.
Then, the number of events that have taken place is Poisson
distributed with rate λ.

Example
Historically, an emergency room after hours (10pm–6am) sees 48
requests every 8 hours. The time until the next patient arrives is
exponentially distributed with that rate. What is the probability that the
next patient arrives in the next 10 minutes? What is the probability
there are 5 patients during the next hour?
48 requests
Answer: Let’s use λ = 8 hours = 6 requests per hour.
The time to the next patient arriving is exponentially distributed.
The number of patients is Poisson distributed.

7 / 10 Chrysafis Vogiatzis Continuous random variables

Example (cont’d)
Example
Historically, an emergency room after hours (10pm–6am) sees 48
requests every 8 hours. The time until the next patient arrives is
exponentially distributed with that rate. What is the probability that the
next patient arrives in the next 10 minutes? What is the probability
there are 5 patients during the next hour?

Answer:
λ = 6 request per hour. Let T be the time to next patient arrival,
and let X be the number of patients during the next hour.
Time to next patient:
1
P(T ≤ 10 mins) = P(T ≤ 1/6 hrs) = 1 − e−6· 6 = 1 − e−1 =
= 0.6321.
Number of patients:
65
P(X = 5) = e−6 · = 0.1606.
5!

8 / 10 Chrysafis Vogiatzis Continuous random variables

Example (cont’d)
Example
Historically, an emergency room after hours (10pm–6am) sees 48
requests every 8 hours. The time until the next patient arrives is
exponentially distributed with that rate. What is the probability that the
next patient arrives in the next 10 minutes? What is the probability
there are 5 patients during the next hour?

Answer:
λ = 6 request per hour. Let T be the time to next patient arrival,
and let X be the number of patients during the next hour.
Time to next patient:
1
P(T ≤ 10 mins) = P(T ≤ 1/6 hrs) = 1 − e−6· 6 = 1 − e−1 =
= 0.6321.
Number of patients:
65
P(X = 5) = e−6 · = 0.1606.
5!

8 / 10 Chrysafis Vogiatzis Continuous random variables

Example (cont’d)
Example
Historically, an emergency room after hours (10pm–6am) sees 48
requests every 8 hours. The time until the next patient arrives is
exponentially distributed with that rate. What is the probability that the
next patient arrives in the next 10 minutes? What is the probability
there are 5 patients during the next hour?

Answer:
λ = 6 request per hour. Let T be the time to next patient arrival,
and let X be the number of patients during the next hour.
Time to next patient:
1
P(T ≤ 10 mins) = P(T ≤ 1/6 hrs) = 1 − e−6· 6 = 1 − e−1 =
= 0.6321.
Number of patients:
65
P(X = 5) = e−6 · = 0.1606.
5!

8 / 10 Chrysafis Vogiatzis Continuous random variables

 Example
A car transmission fails in time that is exponentially distributed with a
rate of 1 every 80,000 miles. What is the probability that the
transmission does not fail within its first 40,000 miles? What is the
probability that the transmission does not fail in the next 40,000 miles
given that it has already been working for 80,000?

Answer: First, estimate the failure rate as λ = 1/80000.

40000
P (X > 40000) = 1 − F (40000) = 1 − (1 − e− 80000 ) = e−0.5 = 0.607.

We are looking for P(X > 120000|X > 80000). Let s = 120000 and
t = 80000. Then:
P(X >s+t)
z }| {
P (X > s + t ∩ X > s)
P (X > s + t|X > s) = =
P (X > s)
P (X > s + t) e−λ(s+t)
= = −λ(s) = e−λt =
P (X > s) e
= P (X > t) = P (X > 40000) = 0.607.

9 / 10 Chrysafis Vogiatzis Continuous random variables

 Example
A car transmission fails in time that is exponentially distributed with a
rate of 1 every 80,000 miles. What is the probability that the
transmission does not fail within its first 40,000 miles? What is the
probability that the transmission does not fail in the next 40,000 miles
given that it has already been working for 80,000?

Answer: First, estimate the failure rate as λ = 1/80000.

40000
P (X > 40000) = 1 − F (40000) = 1 − (1 − e− 80000 ) = e−0.5 = 0.607.

We are looking for P(X > 120000|X > 80000). Let s = 120000 and
t = 80000. Then:
P(X >s+t)
z }| {
P (X > s + t ∩ X > s)
P (X > s + t|X > s) = =
P (X > s)
P (X > s + t) e−λ(s+t)
= = −λ(s) = e−λt =
P (X > s) e
= P (X > t) = P (X > 40000) = 0.607.

9 / 10 Chrysafis Vogiatzis Continuous random variables

 Example
A car transmission fails in time that is exponentially distributed with a
rate of 1 every 80,000 miles. What is the probability that the
transmission does not fail within its first 40,000 miles? What is the
probability that the transmission does not fail in the next 40,000 miles
given that it has already been working for 80,000?

Answer: First, estimate the failure rate as λ = 1/80000.

40000
P (X > 40000) = 1 − F (40000) = 1 − (1 − e− 80000 ) = e−0.5 = 0.607.

We are looking for P(X > 120000|X > 80000). Let s = 120000 and
t = 80000. Then:
P(X >s+t)
z }| {
P (X > s + t ∩ X > s)
P (X > s + t|X > s) = =
P (X > s)
P (X > s + t) e−λ(s+t)
= = −λ(s) = e−λt =
P (X > s) e
= P (X > t) = P (X > 40000) = 0.607.

9 / 10 Chrysafis Vogiatzis Continuous random variables

 Example
A car transmission fails in time that is exponentially distributed with a
rate of 1 every 80,000 miles. What is the probability that the
transmission does not fail within its first 40,000 miles? What is the
probability that the transmission does not fail in the next 40,000 miles
given that it has already been working for 80,000?

Answer: First, estimate the failure rate as λ = 1/80000.

40000
P (X > 40000) = 1 − F (40000) = 1 − (1 − e− 80000 ) = e−0.5 = 0.607.

We are looking for P(X > 120000|X > 80000). Let s = 120000 and
t = 80000. Then:
P(X >s+t)
z }| {
P (X > s + t ∩ X > s)
P (X > s + t|X > s) = =
P (X > s)
P (X > s + t) e−λ(s+t)
= = −λ(s) = e−λt =
P (X > s) e
= P (X > t) = P (X > 40000) = 0.607.

9 / 10 Chrysafis Vogiatzis Continuous random variables

 Example
A car transmission fails in time that is exponentially distributed with a
rate of 1 every 80,000 miles. What is the probability that the
transmission does not fail within its first 40,000 miles? What is the
probability that the transmission does not fail in the next 40,000 miles
given that it has already been working for 80,000?

Answer: First, estimate the failure rate as λ = 1/80000.

40000
P (X > 40000) = 1 − F (40000) = 1 − (1 − e− 80000 ) = e−0.5 = 0.607.

We are looking for P(X > 120000|X > 80000). Let s = 120000 and
t = 80000. Then:
P(X >s+t)
z }| {
P (X > s + t ∩ X > s)
P (X > s + t|X > s) = =
P (X > s)
P (X > s + t) e−λ(s+t)
= = −λ(s) = e−λt =
P (X > s) e
= P (X > t) = P (X > 40000) = 0.607.

9 / 10 Chrysafis Vogiatzis Continuous random variables

 Example
A car transmission fails in time that is exponentially distributed with a
rate of 1 every 80,000 miles. What is the probability that the
transmission does not fail within its first 40,000 miles? What is the
probability that the transmission does not fail in the next 40,000 miles
given that it has already been working for 80,000?

Answer: First, estimate the failure rate as λ = 1/80000.

40000
P (X > 40000) = 1 − F (40000) = 1 − (1 − e− 80000 ) = e−0.5 = 0.607.

We are looking for P(X > 120000|X > 80000). Let s = 120000 and
t = 80000. Then:
P(X >s+t)
z }| {
P (X > s + t ∩ X > s)
P (X > s + t|X > s) = =
P (X > s)
P (X > s + t) e−λ(s+t)
= = −λ(s) = e−λt =
P (X > s) e
= P (X > t) = P (X > 40000) = 0.607.

9 / 10 Chrysafis Vogiatzis Continuous random variables

 Example
A car transmission fails in time that is exponentially distributed with a
rate of 1 every 80,000 miles. What is the probability that the
transmission does not fail within its first 40,000 miles? What is the
probability that the transmission does not fail in the next 40,000 miles
given that it has already been working for 80,000?

Answer: First, estimate the failure rate as λ = 1/80000.

40000
P (X > 40000) = 1 − F (40000) = 1 − (1 − e− 80000 ) = e−0.5 = 0.607.

We are looking for P(X > 120000|X > 80000). Let s = 120000 and
t = 80000. Then:
P(X >s+t)
z }| {
P (X > s + t ∩ X > s)
P (X > s + t|X > s) = =
P (X > s)
P (X > s + t) e−λ(s+t)
= = −λ(s) = e−λt =
P (X > s) e
= P (X > t) = P (X > 40000) = 0.607.

9 / 10 Chrysafis Vogiatzis Continuous random variables

 Example
A car transmission fails in time that is exponentially distributed with a
rate of 1 every 80,000 miles. What is the probability that the
transmission does not fail within its first 40,000 miles? What is the
probability that the transmission does not fail in the next 40,000 miles
given that it has already been working for 80,000?

Answer: First, estimate the failure rate as λ = 1/80000.

40000
P (X > 40000) = 1 − F (40000) = 1 − (1 − e− 80000 ) = e−0.5 = 0.607.

We are looking for P(X > 120000|X > 80000). Let s = 120000 and
t = 80000. Then:
P(X >s+t)
z }| {
P (X > s + t ∩ X > s)
P (X > s + t|X > s) = =
P (X > s)
P (X > s + t) e−λ(s+t)
= = −λ(s) = e−λt =
P (X > s) e
= P (X > t) = P (X > 40000) = 0.607.

9 / 10 Chrysafis Vogiatzis Continuous random variables

 Example
A car transmission fails in time that is exponentially distributed with a
rate of 1 every 80,000 miles. What is the probability that the
transmission does not fail within its first 40,000 miles? What is the
probability that the transmission does not fail in the next 40,000 miles
given that it has already been working for 80,000?

Answer: First, estimate the failure rate as λ = 1/80000.

40000
P (X > 40000) = 1 − F (40000) = 1 − (1 − e− 80000 ) = e−0.5 = 0.607.

We are looking for P(X > 120000|X > 80000). Let s = 120000 and
t = 80000. Then:
P(X >s+t)
z }| {
P (X > s + t ∩ X > s)
P (X > s + t|X > s) = =
P (X > s)
P (X > s + t) e−λ(s+t)
= = −λ(s) = e−λt =
P (X > s) e
= P (X > t) = P (X > 40000) = 0.607.

9 / 10 Chrysafis Vogiatzis Continuous random variables

 Example
A car transmission fails in time that is exponentially distributed with a
rate of 1 every 80,000 miles. What is the probability that the
transmission does not fail within its first 40,000 miles? What is the
probability that the transmission does not fail in the next 40,000 miles
given that it has already been working for 80,000?

Answer: First, estimate the failure rate as λ = 1/80000.

40000
P (X > 40000) = 1 − F (40000) = 1 − (1 − e− 80000 ) = e−0.5 = 0.607.

We are looking for P(X > 120000|X > 80000). Let s = 120000 and
t = 80000. Then:
P(X >s+t)
z }| {
P (X > s + t ∩ X > s)
P (X > s + t|X > s) = =
P (X > s)
P (X > s + t) e−λ(s+t)
= = −λ(s) = e−λt =
P (X > s) e
= P (X > t) = P (X > 40000) = 0.607.

9 / 10 Chrysafis Vogiatzis Continuous random variables

Memorylessness

Definition
A random variable X is said to be memoryless (without memory) if:

P(X > s + t|X > s) = P(X > t).

Think of X as the lifetime of a printer. Then:

the probability it still works after a year, and
the probability it still works in Year 3, assuming it’s survived until
Year 2 are the same, if X is memoryless!

Exponentially distributed random variables are memoryless! They

are the only continuous random variables with that property.

10 / 10 Chrysafis Vogiatzis Continuous random variables

Memorylessness

Definition
A random variable X is said to be memoryless (without memory) if:

P(X > s + t|X > s) = P(X > t).

Think of X as the lifetime of a printer. Then:

the probability it still works after a year, and
the probability it still works in Year 3, assuming it’s survived until
Year 2 are the same, if X is memoryless!

Exponentially distributed random variables are memoryless! They

are the only continuous random variables with that property.

10 / 10 Chrysafis Vogiatzis Continuous random variables

Memorylessness

Definition
A random variable X is said to be memoryless (without memory) if:

P(X > s + t|X > s) = P(X > t).

Think of X as the lifetime of a printer. Then:

the probability it still works after a year, and
the probability it still works in Year 3, assuming it’s survived until
Year 2 are the same, if X is memoryless!

Exponentially distributed random variables are memoryless! They

are the only continuous random variables with that property.

10 / 10 Chrysafis Vogiatzis Continuous random variables

Memorylessness

Definition
A random variable X is said to be memoryless (without memory) if:

P(X > s + t|X > s) = P(X > t).

Think of X as the lifetime of a printer. Then:

the probability it still works after a year, and
the probability it still works in Year 3, assuming it’s survived until
Year 2 are the same, if X is memoryless!

Exponentially distributed random variables are memoryless! They

are the only continuous random variables with that property.

10 / 10 Chrysafis Vogiatzis Continuous random variables