Chap 3 Some Special Distributions

Chapter 3: Some special distributions
Phan Thi Khanh Van
E-mail: khanhvanphan@hcmut.edu.vn
July 2, 2020
(Phan Thi Khanh Van) Chap 3: Some special distributions July 2, 2020 1 / 79
Table of Contents
1 Bernoulli distribution
2 Binomial distribution
3 Geometric distributions
4 Hypergeometric distribution
5 Negative binomial distributions
6 Poisson distributions. Poisson processes
7 Uniform distributions.
8 Normal distributions. Central limit theorems
9 Log-normal distributions
10 Exponential distributions
11 Erlang distributions and Gamma distributions
12 Weibull distributions
13 Beta distributions
Bernoulli distribution
1 Flip a coin. X is the number of heads. P(X = 1) = 0.5 = P(X = 0).
2 A worn machine tool produces 1% defective parts. Let X = number
of defective part produced in one trial. P(X = 1) = 0.01,
P(X = 0) = 0.99
3 Each sample of air has a 10% chance of containing a particular rare
molecule. Let X = the number of air samples that contain the rare
molecule when the next sample is analyzed. P(X = 1) = 0.1.
4 Of all bits transmitted through a digital transmission channel, 10%
are received in error. Let X = the number of bits in error when the
next bit is transmitted. P(X = 1) = 0.1, P(X = 0) = 0.9.
5 Of all patients suffering a particular illness, 35% experience
improvement from a particular medication. A patient is administered
the medication, let X = the number of patients who experience
improvement. P(X = 1) = 0.35, P(X = 0) = 0.75.
Bernoulli distribution
The random variable X is called a Bernoulli random variable if it takes
the value 1 with probability p and the value 0 with probability q = 1 − p.
Mean and Variance

If X is Bernoulli random variable with parameter p,
µ = E (X ) = p, σ 2 = V (X ) = p(1 − p).
Example
A worn machine tool produces 1% defective parts. Let X = number of
defective part produced in one trial. P(X = 1) = 0.01,
P(X = 0) = 0.99.
E (X ) = 0.01
V (X ) = 0.01 × 0.99 = 0.099.
Binomial distribution
1 Flip a coin 3 times. Let X = number of heads obtained.
2 A worn machine tool produces 1% defective parts. Let X = number
of defective parts in the next 25 parts produced.
3 Each sample of air has a 10% chance of containing a particular rare
molecule. Let X = the number of air samples that contain the rare
molecule in the next 18 samples analyzed.
4 Of all bits transmitted through a digital transmission channel, 10%
are received in error. Let X = the number of bits in error in the next
5 bits transmitted.
5 A multiple-choice test contains 10 questions, each with four choices,
and you guess at each question. Let X = the number of questions
answered correctly.
6 In the next 20 births at a hospital, let X = the number of female
births.
Flip 3 coins
Flip 3 coins, consider the number of heads:
x 0 1 2 3
P(X = x) 18 38 38 18
Camera flash tests

The time to recharge the flash is tested in three cell-phone cameras. The
probability that a camera passes the test is 0.8, and the cameras perform
independently. Consider the number of cameras that pass the test.
x 0 1 2 3
P(X = x) 0.008 0.096 0.384 0.512
Binomial distribution.
A random experiment consists of n Bernoulli trials such that
1 The trials are independent.
2 Each trial results in only two possible outcomes, labeled as “success”
and “failure.”
3 The probability of a success in each trial, denoted as p, remains
constant.
The random variable X that equals the number of trials that result in a
success is a binomial random variable with parameters 0 < p < 1 and
n = 1, 2, · · · .
The probability mass function (PMF) of X is
f (x) = Cnx p x (1 − p)n−x , x = 0, 1, , · · · , n
Digital Channel
The chance that a bit transmitted through a digital transmission channel
is received in error is 0.1. Also, assume that the transmission trials are
independent. Let X = the number of bits in error in the next four bits
transmitted. Determine P(X = 2).
Outcome x Outcome x
OOOO 0 EOOO 1
OOOE 1 EOOE 2
OOEO 1 EOEO 2
OOEE 2 EOEE 3
OEOO 1 EEOO 2
OEOE 2 EEOE 3
OEEO 2 EEEO 3
OEEE 3 EEEE 4
P(X = 2) = 6 · P(E ) · P(O)2 = 6 · 0.12 · 0.92 = 0.0486.
2
Mean and Variance
If X is a binomial random variable with parameter p,
µ = E (X ) = np, σ 2 = V (X ) = np(1 − p).
Digital Channel
PMF of X :
X 0 1 2 3 4
P(X = x) 0.6561 0.2916 0.0486 0.0036 0.0001
P
µ = Px.P(X = x) = 0.4.
σ 2 = x 2 .P(X = x) − µ2 = 4 × 0.1 × 0.9 = 0.36
Example
A multiple-choice test contains 25 questions, each with four answers.
Assume that a student just guesses on each question. Find the probability
(a) that the student answers more than 20 questions correctly?
(b) that the student answers fewer than 5 questions correctly?
Let X be the number of questions that the student answered correctly. X

is a binomial random variable with the parameter p = 0.25.
(a) The probability that the student answers more than 20: P(X > 20)
= P(X = 21) + P(X = 22) + P(X = 23) + P(X = 24) + P(X = 25)
= C2521 .0.2521 0.754 +C 22 .0.2522 0.753 +C 23 .0.2523 0.752
25 25
+C2524 .0.2524 0.751 +C 25 .0.2525 = 9.6769.10−10 .
25
(b) The probability that the student answers fewer than 5: P(X < 5)
= P(X = 0) + P(X = 1) + P(X = 2) + P(X = 3) + P(X = 4)
= C250 .0.7525 +C 1 .0.251 0.7524 +C 2 .0.252 0.7523 +C 3 .0.253 0.7522
25 25 25
+C254 .0.254 0.7521 = 0.2137.
Geometric distributions
Digital Channel
The probability that a bit transmitted through a digital transmission
channel is received in error is 0.1. Assume that the transmissions are
independent events, and let the random variable X denote the number of
bits transmitted until the first error.
Then P(X = 5) is the probability that the first four bits are transmitted
correctly and the fifth bit is in error: {OOOOE }, where O denotes an
okay bit.
P(X = 5) = P(OOOOE ) = 0.94 0.1 = 0.066
Note that there is some probability that X will equal any integer value.
Also, if the first trial is a success, X = 1. Therefore, the range of X is
{1, 2, 3, ...}, that is, all positive integers.
In a series of Bernoulli trials (independent trials with constant probability
p of a success), the random variable X that equals the number of trials
until the first success is a geometric random variable with parameter
0 < p < 1 and
f (x) = (1 − p)x−1 p, x = 1, 2, ...
Mean and Variance

If X is a geometric random variable with parameter p,
µ = E (X ) = 1/p, σ 2 = V (X ) = (1 − p)/p 2 .
Example
Assume that each of your calls to a popular radio station has a probability
of 0.02 of connecting, that is, of not obtaining a busy signal. Assume that
your calls are independent.
(a) What is the probability that your first call that connects is your 10th
call?
(b) What is the probability that it requires more than five calls for you to
connect?
(c) What is the mean number of calls needed to connect?
Let X be the number of calls needed to connect to the radio station.

(a) P(X = 10) = 0.989 · 0.02 = 0.0167
(b) P(X > 5) = 1 − P(X < 5)
= 1 − P(X = 1) − P(X = 2) − P(X = 3) − P(X = 4) − P(X = 5)
= 1 − 0.02 − 0.98 · 0.02 − 0.982 · 0.02 − 0.983 · 0.02 − 0.984 · 0.02 = 0.9039.
(c) µ(X ) = 1/0.02 = 50.
Example
An array of 30 LED bulbs is used in an automotive light. The probability
that a bulb is defective is 0.001 and defective bulbs occur independently.
Determine the following:
a) Probability that an automotive light has two or more defective bulbs.
b) Expected number of automotive lights to check to obtain one with
two or more defective bulbs.
a) Let X be the number of defective bulbs. X is a binomial random
variable.
P(X ≥ 2) = 1 − P(X = 0) − P(X = 1) =
0 0.99930 − C 1 0.001 · 0.99929 ≈ 0.000423
1 − C30 30
b) Let Y be the number of automotive lights to check to obtain one
with two or more defective bulbs. Y is a geometric random variable
with the parameter p = P(X ≥ 2) = 0.000423. Therefore, the
expected number of automotive lights to check to obtain one with
two or more defective bulbs: E (Y ) = 1/p ≈ 2364 (lights).
Hypergeometric distribution
Parts from Suppliers

A batch of parts contains 100 from a local supplier of tubing and 200 from
a supplier of tubing in the next state. If four parts are selected randomly
and without replacement
(a) what is the probability they are all from the local supplier?
(b) what is the probability that two or more parts in the sample are from
the local supplier?
(c) What is the probability that at least one part in the sample is from the
local supplier?
Let X equal the number of parts in the sample from the local supplier.
C 4 .C200
0
(a) P(X = 4) = 100 4
C300
= 0.0119.
(b) P(X ≥ 2) = P(X = 2) + P(X = 3) + P(X = 4) = 0.408.
(c) P(X ≥ 1) = 1 − P(X = 0) = 0.804.
A set of N objects contains K objects classified as successes and N − K
objects classified as failures. A sample of size n objects is selected
randomly (without replacement) from the N objects where K ≤ N and
n ≤ N. The random variable X that equals the number of successes in the
sample is a hypergeometric random variable and
n−x
CKx .CN−K
f (x) = , x = max{0, n + K − N} to min{K , n}.
CNn
Mean and Variance
If X is a hypergeometric random variable with parameters N, K , and n,
then
µ = E (X ) = np and
σ 2 = V (X ) = np(1 − p) N−n
N−1 , where p = K /N.
Finite population correction factor

The term in the variance of a hypergeometric random variable
N−n
N−1
is called the finite population correction factor
If sampling were done with replacement, X would be a binomial

random variable and its variance would be np(1 − p).
If n N, the correction is small and the hypergeometric distribution
is similar to the binomial distribution ⇒ a binomial distribution can
effectively approximate the hypergeometric distribution.
Example
A state runs a lottery in which 6 numbers are randomly selected from 40
without replacement. A player chooses 6 numbers before the state’s
sample is selected.
a) What is the probability that the 6 numbers chosen by a player match
all 6 numbers in the state’s sample?
b) What is the probability that 5 of the 6 numbers chosen by a player
appear in the state’s sample?
c) What is the probability that 4 of the 6 numbers chosen by a player
appear in the state’s sample?
d) If a player enters one lottery each week, what is the expected number
of weeks until a player matches all 6 numbers in the state’s sample?
Let X be the number of numbers chosen by a player that match the
numbers in the state’s sample. X is a hypergeometric random variable.
C66 C34
0
a) P(X = 6) = C406 = 2.6053.10−7
C65 C34
1
b) P(X = 5) = C406
C64 C34
2
c) P(X = 4) = C406
d) Assume that the ”success” is the event that the player matches all 6
numbers in the state’s sample.
Let Y be number of weeks until the first ”success”. Then Y is a
geometric random variable with the parameter
p = P(X = 6) = 2.6053.10−7 .
The expected number of weeks until a player matches all 6 numbers
in the state’s sample: E (Y ) = 1/p = 3838380.
Negative Binomial Distribution
Camera flashes
Consider the time to recharge the flash of cameras. The probability that a
camera passes the test is 0.8, and the cameras perform independently.
What is the probability that the third failure is obtained in five or fewer
tests?
Let X denote the number of cameras tested until 3 failures have been
obtained.
The requested probability is P(X ≤ 5). Here X has a negative binomial
distribution with p = 0.2 and r = 3.
Therefore,
P(X ≤ 5) = P(X = 3) + P(X = 4) + P(X = 5)
= C22 0.23 + C32 0.23 · 0.8 + C42 0.23 · 0.82 = 0.056.
Negative Binomial Distribution
In a series of Bernoulli trials (independent trials with constant probability
p of a success), the random variable X that equals the number of trials
until r successes occur is a negative binomial random variable with
parameters 0 < p < 1 and r = 1, 2, 3, ..., and
r −1
f (x) = Cx−1 (1 − p)x−r p r , x = r , r + 1, r + 2, ...
Mean and Variance

If X is a negative binomial random variable with parameters p and r ,
µ = E (X ) = r /p, σ 2 = V (X ) = r (1 − p)/p 2 .
Example
In a clinical study, volunteers are tested for a gene that has been found to
increase the risk for a disease. The probability that a person carries the
gene is 0.1.
a) What is the probability that four or more people need to be tested to
detect two with the gene?
b) What is the expected number of people to test to detect two with the
gene?
(a) Let X be the number of people needed to be tested to detect 2 with

the gene. X is a negative binomial random variable, p = 0.1, r = 2.
P(X ≥ 4) = 1 − P(X = 2) − P(X = 3) = 1 − 0.12 − C21 0.12 · 0.9 = 0.972
(b) E (X ) = pr = 0.1
2
= 20.
Random processes
In real-life applications, we are often interested in multiple observations of

random values over a period of time. For example S(t) - stock price at the
time t ∈ [0, ∞). At any fixed time t∗ , S(t∗ ) is a random variable.
S(t), t ∈ [0, ∞) - random process or stochastic process.
Random processes
A random process is a collection of random variables usually indexed by
time.
Continuous-time and discrete-time random processes
A continuous-time random process is a random process
{X (t), t ∈ J}, where J is an interval on the real line such as [−1, 1],
[0, ∞), (−∞, ∞), etc.
A discrete-time random process (or a random sequence) is a random
process {X (n) = Xn , n ∈ J}, where J is a countable set such as N or
Z.
Let N(t) ∈ {0, 1, 2...} be the number of customers who have visited a
bank from t = 9 am until time t, on a given day, for t ∈ [9, 16] -
continuous-time.
Let W (t) ∈ R be the thermal noise voltage generated across a resistor
in an electric circuit at time t, for t ∈ [0, ∞) - continuous-time.
Let X (tn ) = Xn ∈ [25, 39] be the temperature in HCM city in the
n − th day of May, n = 1, 2, 3....31 - discrete-time.
Poisson processes
The Poisson process is one of the most widely-used counting processes. It

is usually used in scenarios where we are counting the occurrences of
certain events that appear to happen at a certain rate, but completely at
random.
For example:
earthquakes occur in a certain area with a rate of λ = 2 per month
and the timings of earthquakes seem to be completely random.
the number of car accidents at a site or in an area;
the location of users in a wireless network;
the requests for individual documents on a web server;
the outbreak of wars;
photons landing on a photodiode.
Poisson random variable
The random variable X that equals the number of events that occur in the
interval of length T in a Poisson process is a Poisson random variable
with parameter 0 < λ, and
e −λT (λT )x
f (x) = , x = 0, 1, 2...
x!
Mean and Variance

If X is a Poisson random variable over an interval of length T with
parameter λ, then
µ = E (X ) = λT , σ 2 = V (X ) = λT .
Example 1
The number of telephone calls that arrive at a phone exchange is often
modeled as a Poisson random variable. Assume that on the average there
are 10 calls per hour.
a) What is the probability that there are exactly 5 calls in one hour?
b) What is the probability that there are 3 or fewer calls in one hour?
c) What is the probability that there are exactly 15 calls in two hours?
d) What is the probability that there are exactly 5 calls in 30 minutes?
Let X , Y , Z be the number of calls that arrive at a phone exchange in

1, 2, 0.5 hours, respectively.
e −10 105
a) λ = 10, T = 1: P(X = 5) = 5! = 0.0378.
b) P(X ≤ 3) = P(X = 0) + P(X = 1) + P(X = 2) + P(X = 3)
−10 100 101 102 103
=e 0! + 1! + 2! + 3! = 0.0103.
e −20 2015
c) T = 2, P(Y = 15) = 15! ≈ 0.0516.
e −5 55
d) T = 0.5, P(Z = 5) = 5! ≈ 0.1755.
Example 2
Astronomers treat the number of stars in a given volume of space as a
Poisson random variable. The density in the Milky Way Galaxy in the
vicinity of our solar system is one star per 16 cubic light-years.
a) What is the probability of two or more stars in 16 cubic light-years?
b) How many cubic light-years of space must be studied so that the
probability of one or more stars exceeds 0.95?
a) Let X be the number of stars in 16 cubic light-years. λ = 1.

In 16 cubic light-years: T = 1: P(X ≥ 2)
−1 0 −1
1 − P(X = 0) − P(X = 1) = 1 − e 0!1 − e1! = 0.2642.
b) Let N be the number of cubic light-years must be studied and X be
N
the number of stars in N cubic light-years. We have: T = 16 .
We must have P(Y ≥ 1) > 0.95 ⇔ 1 − P(Y = 0) > 0.95
−N
⇔ P(Y = 0) = e −1.T = e 16 ≤ 0.05
⇔ N ≥ −16 ln 0.05 = 47.9317 (cubic light-years).
Example 3
The number of customers arriving at a grocery store can be modeled by a
Poisson process with intensity λ = 10 customers per hour.
a) Find the probability that there are 2 customers between 10 : 00 and
10 : 20.
b) Find the probability that there are 3 customers between 10 : 00 and
10 : 20 and 7 customers between 10 : 20 and 11 : 00.
λ = 10, The interval I1 : 10 : 00 − 10 : 20, I2 : 10 : 20 − 11 : 00. The

length of these interval T1 = 13 , T2 = 32 hours. Let X and Y be the
number of customers arriving in the intervals I1 and I2 , respectively. X , Y
are a Poisson random variables.
−10
e 3 ( 10 )2
a) Therefore, P(X = 2) = 3
2! = 0.1982
b) Because I1 and I2 are two disjoint intervals, we have:
P(3 arrive in I1 , 7 arrive in I2 ) = P(3 arrive in I1 ).P(7 arrive in I2 )
−10 −20
e 3 ( 10 )3 e 3 ( 20 )7
= P(X = 3).P(Y = 7) = 3
3! . 7!
3
= 0.0325.
Uniform distributions.
Uniform distributions
A continuous uniform random variable X is a continuous random
variable with: (
1
, a ≤ x ≤ b,
PDF: f (x) = b−a .
0, x < a or x > b

0
 x < a,
x−a
CDF: F (x) = b−a , a ≤ x ≤ b,

1, x >b

Mean and Variance

If X is a continuous uniform random variable over a ≤ x ≤ b
(a+b) (b−a)2
µ = E (x) = 2 , and σ 2 = 12
Example 1
An e-mail message will arrive at a time uniformly distributed between
9 : 00 a.m. and 11 : 00 a.m. You check e-mail at 9 : 15 a.m. and every 30
minutes afterward.
a) What is the standard deviation of arrival time (in minutes)?
b) What is the probability that the message arrives less than 10 minutes
before you view it?
c) What is the probability that the message arrives more than 15
minutes before you view it?
Let X be the time

qthat an email message arrive.
√ (11−9) 2
q
1
a) σ= V = 12 = 3 (hour).
1
b) PDF: f (x) = b−a = 0.5 The probability that the message arrives less
than 10 minutes before you view it:
P(9 : 05 ≤ X ≤ 9 : 15) + P(9 : 35 ≤ X ≤ 9 : 45)
+P(10 : 05 ≤ X ≤ 10 : 15) + P(10 : 35 ≤ X ≤ 10 : 45)
= 4 · 0.5 · 16 ) = 0.3333
c) The probability that the message arrives more than 15 minutes before
you view it:
P(9 : 15 ≤ X ≤ 9 : 30) + P(9 : 45 ≤ X ≤ 10 : 00) + P(10 : 15 ≤ X ≤
10 : 30) = 0.5 · 3 · 41 = 0.375
Example 2
The volume of a shampoo filled into a container is uniformly distributed
between 374 and 380 milliliters.
a) What are the mean and standard deviation of the volume of
shampoo?
b) What is the probability that the container is filled with less than the
advertised target of 375 milliliters?
c) What is the volume of shampoo that is exceeded by 95% of the
containers?
d) Every milliliter of shampoo costs the producer $0.002. Any shampoo
more than 375 milliliters in the container is an extra cost to the
producer. What is the mean extra cost?
a) Let X be the volume of a shampoo filled into a container. X is a
uniform random variable with a = 374, b = 380.
E (X ) = a+b = 377
p 2 b−a
√
σ = V (x) = √ 12
= 3.
R 375
b) P(X < 375) = 374 dt 1
6 = 6.
R 380
c) P(X > x0 ) = 95% ⇔ x0 dt 6 = 0.95
380−x0
⇔ 6 = 0.95 ⇔ x0 = 380 − 6 · 0.95 = 374.3.
d) The mean extra cost:
C = (E (X ) − 375) · 0.002 = ( 374+380
2 − 375) · 0.002 = 0.004($)
Normal distributions
A random variable X with probability density function
−(x−µ)2
f (x) = √1 e 2σ 2 , −∞ < x < ∞
2πσ
is a normal random variable with parameters µ where −∞ < µ < ∞,

and σ > 0. Also,
E (X ) = µ and V (X ) = σ 2 .
and the notation N(µ, σ 2 ) is used to denote the distribution.
For any normal random variable:
P(µ − σ < X < µ + σ) = 0.68
P(µ − 2σ < X < µ + 2σ) = 0.95
P(µ − 3σ < X < µ + 3σ) = 0.997
Example 1
Assume that the current measurements in a strip of wire follow a normal
distribution with a mean of 10 milliamperes and a variance of 4
(milliamperes) . What is the probability that a measurement exceeds 13
milliamperes?
Let X denote the current in milliamperes.

R∞ 1 −(x−10)2
P(X > 13) = √2π.2 e 2.4 dx: We often use table to evaluate the
13
integral.
Standard normal random variable
A normal random variable X with
µ = 0, σ2 = 1
is called a standard normal random variable and is denoted as Z .

The CDF of a standard normal random variable is denoted as
Φ(Z ) = P(Z ≤ z).
Standardizing a Normal Random Variable

If X is a normal random variable with
−(x−µ)2
E (X ) = µ, V (X ) = σ 2 , PDF: f (x) = √1 e 2σ 2 ,
2πσ
X −µ
the random variable Z = σ is a normal random variable with
E (Z ) = 0, V (Z ) = 1 : Z ∼ N(0, 1) : standard normal random
variable.
Example 2
The time until recharge for a battery in a laptop computer under common
conditions is normally distributed with a mean of 260 minutes and a
standard deviation of 50 minutes. What is the probability that a battery
lasts more than four hours?
Let X be the time until recharge for a battery in a laptop computer under
common conditions. X ∼ N(260, 502 )
Change the variable: Z = X σ−µ = X −260
50 , we have Z ∼ N(0, 1)
R∞ 1 −z 2
P(X ≥ 240) = P(Z ≥ −0.4) = √ e 2 dz = 0.6554
2π
−0.4
Or we can use the table: P(Z ≥ −0.4) = 1 − P(Z < −0.4) = 1 − 0.3446.
Example 3
The weight of a sophisticated running shoe is normally distributed with a
mean of 12 ounces and a standard deviation of 0.5 ounce.
a) What is the probability that a shoe weighs more than 13 ounces?
b) What value of weight is exceeded with 95% probability?
c) What must the standard deviation of weight be in order for the
company to state that 99.9% of its shoes weighs less than 13 ounces?
d) If the standard deviation remains at 0.5 ounce, what must the mean
weight be for the company to state that 99.9% of its shoes weighs less
than 13 ounces?
Let X be the weight of a sophisticated running shoe.

σ = 0.5, µ = 12.
−12
Change the variable: Z = X0.5 we have: Z ∼ N(0, 1).
a) X = 13 ⇒ Z = 2.
P(Z > 2) = 1 − P(Z ≤ 2) (use table) = 1 − 0.9773 = 0.0227
b) P(Z > z0 ) = 0.95 ⇒ P(Z ≤ z0 ) = 0.05 (use table) ⇒ z0 = −1.64
⇒ x0 = z0 · 0.5 + 12 = 11.18
c) In order for the company to state that 99.9% of its shoes weighs less
than X = 13 ounces: (use table) Z = 3.1
⇒ σ = X Z−µ = 13−12
3.1 = 0.3226.
d) If σ = 0.5, then µ = X − Z σ = 13 − 3.1 · 0.5 = 11.45.
Central limit theorem
Let X1 , X2 , · · · be a sequence of independent and identically distributed
(i.i.d.) random variables, each having mean µ and variance σ 2 .
Then the distribution of
X1 + ....Xn − nµ
√
σ n
tends to the standard normal N(0, 1) as n → ∞. That is, for
−∞ < a < ∞,
n o Ra −x 2 /2
√ n −nµ ≤ a → √1
P X1 +···+X e dx, as n → ∞
σ n 2π
−∞
Normal Approximation to the Binomial Distribution
Example
Assume that in a digital communication channel, the number of bits
received in error can be modeled by a binomial random variable, and
assume that the probability that a bit is received in error is 10−5 . If 16
million bits are transmitted, what is the probability that 150 or fewer errors
occur?
Let the random variable X denote the number of errors. Then X is a

binomial random variable and
150
x (10−5 )x (1 − 10−5 )16,000,000−x
P
P(X ≤ 150) = C16,000,000
x=0
- this probability is difficult to compute. The normal distribution can be

used to provide an excellent approximation in this example.
Normal Approximation to the Binomial Distribution
If X is a binomial random variable with parameters n and p,
Z = √X −np
np(1−p)
is approximately a standard normal random variable. To approximate a

binomial probability with a normal distribution, a continuity correction is
applied as follows:

x+0.5−np
P(X ≤ x) = P(X ≤ x + 0.5) ≈ P Z ≤ √
np(1−p)
and

x−0.5−np
√
P(X ≥ x) = P(X ≥ x − 0.5) ≈ P Z ≥
np(1−p)
The approximation is good for np > 5 and n(1 − p) > 5.
Return to the previous example
Assume that in a digital communication channel, the number of bits
received in error can be modeled by a binomial random variable, and
assume that the probability that a bit is received in error is 10−5 . If 16
million bits are transmitted, what is the probability that 150 or fewer errors
occur?
Let the random variable X denote the number of errors. Then X is a

binomial random variable and

X −160 150.5−160
P(X ≤ 150) = P(X ≤ 150.5) = P √ −5
≤√ −5
≈
160(1−10 ) 160(1−10 )
P (Z ≤ −0.75) = 0.227.
Here, np = 16, 000, 000 · 10−5 = 160, n(1 − p) > 5.
Normal Approximation to Poisson Distribution
If X is a Poisson random variable with E (X ) = λ and V (X ) = λ,
X√−λ
Z= λ
is approximately a standard normal random variable. To approximate a

binomial probability with a normal distribution, a continuity correction is
applied as follows:

P(X ≤ x) = P(X ≤ x + 0.5)Z ≈ P Z ≤ x+0.5−λ √
λ
and

x−0.5−λ
P(X ≥ x) = P(X ≥ x − 0.5)Z ≈ P Z ≥ √
λ
The approximation is good for λ > 5.
Example 1
The number of students who enroll in a psychology course is a Poisson
random variable with mean µ = 100. The professor in charge of the course
has decided that if the number enrolling is 120 or more, he will teach the
course in two separate sections, whereas if fewer than 120 students enroll,
he will teach all of the students together in a single section. What is the
probability that the professor will have to teach two sections?
∞
100i
The exact solution: e −100
P
i! : difficult to compute!
i=120
If X denotes the number of students who enroll in the course, we have
P(X ≥ 120) = P(X ≥ 119.5) (the continuity correction)

= P( X√−100
100
≥ 119.5−100
√
100
) ≈ 1 − Φ(1.95) = 0.0256
Example 2
Assume that the number of asbestos particles in a squared meter of dust
on a surface follows a Poisson distribution with a mean of 1000. If a
squared meter of dust is analyzed, what is the probability that 950 or
fewer particles are found?
Let the random variable X denote the number of asbestos particles in a

squared meter of dust on the surface.
950
P e −1000 1000x
P(X ≤ 950) = x!
x=0
The probability can be approximated as

P(X ≤ 950) = P(X ≤ 950.5)
950.5−1000
≈ P Z ≤ √1000 = P(Z ≤ −1.57) = 0.058.
Log-normal distributions
Variables in a system sometimes follow an exponential relationship as

x = exp(w ). If the exponent is a random variable W, then X = exp(W ) is
a random variable with a distribution of interest.
Log-normal distributions
Let W have a normal distribution with mean θ and variance ω 2 ; then
X = exp(W ) is a log-normal random variable with probability density
function
(ln(x) − θ)2

1
f (x) = xω√2π exp − 0<x <∞
2ω 2
The mean and variance of X are
2 /2 2
2

E (X ) = e θ+ω and V (X ) = e 2θ+ω eω − 1
Example
Suppose that the length of stay (in hours) at a hospital emergency
department is modeled with a lognormal random variable X with θ = 1.5
and ω = 0.4. Determine the following in parts (a) and (b):
a) Mean and variance
b) P(X < 8)
2 2
a) Mean E (X ) = e θ+ω /2= e 1.5+0.4

/2 = 4.855.

2 2 2 2
Variance V (X ) = e 2θ+ω e ω − 1 = e 3+0.4 e 0.4 − 1 = 4.0898.
R8 (ln x − 1.5)2

1√
b) P(X < 8) = x·0.4· 2π exp − dx = 0.9263
0 2 · 0.42
Exponential Distribution
The random variable X that equals the distance between successive events
from a Poisson process with mean number of events λ > 0 per unit interval
is an exponential random variable with parameter λ. The PDF of X is
PDF: f (x) = λe −λx , for 0 ≤ x < ∞

Rx
CDF: F (x) = 0 f (t)dt = 1 − e −λx .
Mean and variance

If the random variable X has exponential distribution with parameter λ,
µ = E (X ) = λ1 , and σ 2 = V (X ) = 1
λ2
Memoryless
If the random variable X has exponential distribution with parameter λ,
then X is memoryless random variable, that is

P(X > x + aX > a) = P(X > x), for a, x ≥ 0
Computer Usage
In a large corporate computer network, user log-ons to the system can be
modeled as a Poisson process with a mean of 25 log-ons per hour. What is
the probability that there are no log-ons in an interval of six minutes?
Let X denote the time in hours from the start of the interval until the first
log-on. Then X has an exponential distribution with λ = 25 log-ons per
hour.
We are interested in the probability that X exceeds 6 minutes. Because λ
is given in log-ons per hour, we express all time units in hours. That is, 6
minutes = 0.1 hour.
R∞
Therefore, P(X > 0.1) = 25e −25x dx = e −25(0.1) = 0.082.
0.1
The CDF also can be used to obtain the same result as follows:
P(X > 0.1) = 1 − F (0.1) = e −25(0.1) = e −2.5 .
Example 2
Web crawlers need to estimate the frequency of changes to Web sites to
maintain a current index for Web searches. Assume that the changes to a
Web site follow a Poisson process with a mean of 3.5 days.
a) What is the probability that the next change occurs in less than 2.0
days?
b) What is the probability that the time until the next change is greater
7.0 days?
c) What is the time of the next change that is exceeded with probability
90%
d) What is the probability that the next change occurs in less than 10.0
days, given that it has not yet occurred after 3.0 days?
Let X denote the time in days from 2 changes of Web sites. X is an

x
1 1 − 3.5
exponential random variable with λ = 3.5 . PDF: f (x) = 3.5 e .
Rx t −x
1 − 3.5
CDF: F (x) = 0 3.5 e dt = 1 − e 3.5
a) P(X < 2) = F (2) = 0.4353.
b) P(X > 7) = 1 − P(X < 7) = 1 − F (7) = 0.1353
−x0
c) P(X > x0 ) = 1 − P(X < x0 ) = e 3.5 = 0.9
⇒ x0 = −3.5 ln 0.9 = 0.3688
R 10 t
1 − 3.5
P(3<X <10) 3 3.5 e dt
d) P(X < 10|X > 3) = P(X >3) = R∞ 1 − t = 0.8647
3 3.5 e
3.5 dt
F (10) − F (3)
Or we can use: P(X < 10|X > 3) = = 0.8647.
1 − F (3)
Another method: Using the memoryless property of the exponential
random variable:
7
P(X < 10|X > 3) = P(X < (10 − 3)) = F (7) = 1 − e − 3.5 = 0.8647.
Erlang distributions and Gamma distributions
Processor Failure
The failures of the central processor units of large computer systems are
often modeled as a Poisson process. Typically, failures are not caused by
components wearing out but by more random failures of the large number
of semiconductor circuits in the units. Assume that the units that fail are
immediately repaired, and assume that the mean number of failures per
hour is 0.0001. Let X denote the time until four failures occur in a
system. Determine the probability that X exceeds 40, 000 hours.
Let the random variable N denote the number of failures in 40, 000 hours
of operation. The time until four failures occur exceeds 40, 000 hours if
and only if the number of failures in 40, 000 hours is three or less.
Therefore,
P(X > 40, 000) = P(N ≤ 3)
The assumption that the failures follow a Poisson process implies that N
has a Poisson distribution with
E (N) = 40, 000(0.0001) = 4 failures per 40, 000 hours
Therefore,
3
P e −4 4k
P(X > 40, 000) = P(N ≤ 3) = k! = 0.433
k=0
Gamma functions
The gamma function is
R∞
Γ(r ) = x r −1 e −x dx, for r > 0.
0
Gamma distributions
The random variable X with PDF
λr x r −1 e −λx
f (x) = Γ(r ) , for x > 0
is a gamma random variable with parameter λ > 0 and r > 0. If r is an

integer, X has an Erlang distribution.
Mean and Variance

If X is a gamma random variable with parameters λ and r ,
µ = E (x) = r /λ, and σ 2 = V (X ) = r /λ2 .
Gamma distributions
Example 1
The time between failures of a laser in a cytogenics machine is
exponentially distributed with a mean of 25, 000 hours.
(a) What is the expected time until the second failure?
(b) What is the probability that the time until the third failure exceeds
50, 000 hours?
a) Let X be the time until the second failure. X is an Erlang random

1
variable with r = 2, λ = 25000 = 0.00004.
r
E (X ) = λ = 50000 (hours).
b) Let Y be the time until the third failure. Y is an Erlang random
variable with r = 3, λ = 0.00004.
R∞ λr x r −1 e −λx R∞ 0.000043 ·x 2 e −0.00004x
P(Y > 50000) = Γ(r ) = Γ(3) dx = 0.6767
50000 50000
Gamma distributions
Example 2
Raw materials are studied for contamination. Suppose that the number of
particles of contamination per pound of material is a Poisson random
variable with a mean of 0.01 particle per pound.
(a) What is the expected number of pounds of material required to obtain
15 particles of contamination?
(b) What is the standard deviation of the pounds of materials required to
obtain 15 particles of contamination?
Let X be the number of pounds of material required to obtain 15 particles

of contamination. X is a Erlang random variable with r = 15, λ = 0.01.
a) E (X ) = λr = √0.01
15
= 1500.
15
pr
b) σ = λ2 = 0.01 = 387.2983.
Weibull distributions
The Weibull distribution is often used to model the time until failure of
many different physical systems: bearing wear, some
semiconductorsfailures caused by external shocks to the system.
The random variable X with PDF
β−1 h i
x β
f (x) = βδ xδ

exp − δ , for x > 0
is a Weibull random variable with scale parameter δ > 0 and shape

parameter β > 0.
CDF, Mean and Variance
If X is a Weibull random variable with parameters δ and β,
β
−( δx )
CDF: F (X ) = 1 − e .
Mean: µ = E (x) = δΓ 1 + β1 .
h i2
Variance: σ 2 = V (X ) = δ 2 Γ 1 + β2 − δ 2 Γ 1 + β1 .
Example
Assume that the life of a packaged magnetic disk exposed to corrosive
gases has a Weibull distribution with β = 0.5 and the mean life is 600
hours. Determine the following:
a) Probability that a disk lasts at least 500 hours.
b) Probability that a disk fails before 400 hours.
Let X be the life of a packaged magnetic disk exposed to corrosive gases.

µ = δΓ(1 + β1 ) = 600
600
⇒ δ = Γ(1+2) = 300.
0.5

−( 500
300 )
a) P(X ≥ 500) = 1 − P(X

< 500)
=e = 0.275.
0.5
−( 400
300 )
b) P(X < 400) = 1 − e = 0.6848.
Beta distributions
A continuous distribution that is flexible but bounded over a finite range is
useful for probability models. For example: The proportion of solar
radiation (∈ [0, 1]) absorbed by a material or the proportion (of the
maximum time) required to complete a task in a project.
Beta distributions
The random variable X with probability density function
Γ(α + β) α−1
f (x) = x (1 − x)β−1 , for x in [0, 1]
Γ(α)Γ(β)
is a beta random variable with parameters α > 0 and β > 0.
Beta distributions
If X has a beta distribution with parameters α and β
α αβ
E (X ) = α+β and V (X ) = (α+β)2 (α+β+1)
Beta distributions
Example 1
The maximum time to complete a task in a project is 2.5 days. Suppose
that the completion time as a proportion of this maximum is a beta
random variable with α = 2 and β = 3. What is the probability that the
task requires more than two days to complete?
Let X be the proportion of the maximum time to complete a task in that

2
project: X ∈ [0, 1]. The task requires 2 days to complete: X = 2.5 = 0.8
1
R Γ(5) 1
R 24
2 2
P(X > 0.8) = Γ(2)Γ(3) x(1 − x) dx = 1·2 x(1 − x) dx = 0.0272
0.8 0.8
Beta distributions
Example 2
An allele is an alternate form of a gene, and the proportion of alleles in a
population is of interest in genetics. An article in BMC Genetics
[“Calculating Expected DNA Remnants From Ancient Founding Events in
Human Population Genetics” (2008, Vol. 9:66)] used a beta distribution
with mean 0.3 and standard deviation 0.17 to model initial allele
proportions in a genetic simulation. Determine the parameters α and β for
this beta distribution.

µ = α = 0.3,
(
α+β α = 73 β,
We have q
αβ ⇔ 1.47
10β+7 = 0.0289
σ = = 0.17
(α+β)2 (α+β+1)
(
α = 1.8799,
⇔
β = 4.3865
Thank you for your attention!

Chap 3 Some Special Distributions

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chap 3 Some Special Distributions

Uploaded by

Copyright:

Available Formats

Chapter 3: Some special distributions

Phan Thi Khanh Van

Mean and Variance

Camera flash tests

f (x) = Cnx p x (1 − p)n−x , x = 0, 1, , · · · , n

µ = E (X ) = np, σ 2 = V (X ) = np(1 − p).

Let X be the number of questions that the student answered correctly. X

f (x) = (1 − p)x−1 p, x = 1, 2, ...

Mean and Variance

Let X be the number of calls needed to connect to the radio station.

Parts from Suppliers

Finite population correction factor

is called the finite population correction factor

If sampling were done with replacement, X would be a binomial

Mean and Variance

(a) Let X be the number of people needed to be tested to detect 2 with

In real-life applications, we are often interested in multiple observations of

The Poisson process is one of the most widely-used counting processes. It

Mean and Variance

Let X , Y , Z be the number of calls that arrive at a phone exchange in

a) Let X be the number of stars in 16 cubic light-years. λ = 1.

λ = 10, The interval I1 : 10 : 00 − 10 : 20, I2 : 10 : 20 − 11 : 00. The

Mean and Variance

Let X be the time

is a normal random variable with parameters µ where −∞ < µ < ∞,

and the notation N(µ, σ 2 ) is used to denote the distribution.

Let X denote the current in milliamperes.

is called a standard normal random variable and is denoted as Z .

Φ(Z ) = P(Z ≤ z).

Standardizing a Normal Random Variable

Let X be the weight of a sophisticated running shoe.

Let the random variable X denote the number of errors. Then X is a

- this probability is difficult to compute. The normal distribution can be

is approximately a standard normal random variable. To approximate a

The approximation is good for np > 5 and n(1 − p) > 5.

Let the random variable X denote the number of errors. Then X is a

Here, np = 16, 000, 000 · 10−5 = 160, n(1 − p) > 5.

is approximately a standard normal random variable. To approximate a

The approximation is good for λ > 5.

P(X ≥ 120) = P(X ≥ 119.5) (the continuity correction)

Let the random variable X denote the number of asbestos particles in a

The probability can be approximated as

Variables in a system sometimes follow an exponential relationship as

PDF: f (x) = λe −λx , for 0 ≤ x < ∞

Mean and variance

Let X denote the time in days from 2 changes of Web sites. X is an

P(X > 40, 000) = P(N ≤ 3)

E (N) = 40, 000(0.0001) = 4 failures per 40, 000 hours

is a gamma random variable with parameter λ > 0 and r > 0. If r is an

Mean and Variance

µ = E (x) = r /λ, and σ 2 = V (X ) = r /λ2 .

a) Let X be the time until the second failure. X is an Erlang random

Let X be the number of pounds of material required to obtain 15 particles

is a Weibull random variable with scale parameter δ > 0 and shape

Let X be the life of a packaged magnetic disk exposed to corrosive gases.

Let X be the proportion of the maximum time to complete a task in that

You might also like