CONTINUOUS PROBABILITY DISTRIBUTIONS II

Dr. J. Jagan Mohan

THE EXPONENTIAL DISTRIBUTION
 A continuous random variable X is said to have an
Exponential distribution with parameter λ (λ > 0) if
and only if its pdf is given by

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 e   x , x  0,
f ( x;  )  
 0, otherwise.

 The cdf of Exponential distribution is given by

 0, x  0,
F ( x)    x
1  e , x  0.
2
 Dr Jagan Mohan Jonnalagadda, BITS
Pilani, Hyderabad Campus.
3
EXPONENTIAL DENSITY CURVES
 PROPERTIES
 If X has the Exponential distribution, then

1
  E( X )  .

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS


1
  V(X) 
2
.
 2


M X (t )  , t  .
 t

4
COMPARING GEOMETRIC (DISCRETE) &
EXPONENTIAL (CONTINUOUS) DISTRIBUTIONS:
 In continuous case, exponential function = eλx, λ and x
are real numbers.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 In discrete case, exponential function = qx, q is a real
number and x is an integer.

 PDF of exponential distribution = λeλx, x ≥ 0.

 PMF of geometric distribution = pqx, x is a whole

number. Here q = (1 – p).
5
 Consider a Poisson process with parameter λ.
 Let W denote the time of occurrence of the first event.
The distribution function F of W is given by

F ( w)  P (W  w)  1  P (W  w).

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 Here P(W > w) = The probability of first occurrence of
the event after time w = The probability of no
occurrences of the event in the time interval [0, w].
 Let X denote the number of occurrences of the event in
the time interval [0, w]. Then, X ~ p(λw) and hence

F ( w)  1  P (W  w)  1  e   w .
 Consequently,
w 6
f ( w)  F ( w)   e , w  0.
RELATIONSHIP TO THE POISSON PROCESS
 Suppose that the number of events occurring in any
time interval of length w has a Poisson distribution
with parameter λw (where α is the expected number

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
of events occurring in unit of time).

 Then, the distribution of elapsed time between the

occurrence of two successive events is exponential
with parameter λ.

7
MEMORYLESS PROPERTY
 A non negative random variable X is memoryless if

P ( X  s  t | X  t )  P ( X  s ) for all s, t  0.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 If X has the Exponential distribution, then X is
memoryless.

 In discrete case, if X has the Geometric distribution,

then X is memoryless.

8
 Let X be the life time of some instrument. Then, the
probability that the instrument survives at least (s + t)
hours, given that it has survived t hours, is same as the
initial probability that it survive for at least s hours.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 In other words, if the instrument is alive at age t, the
distribution of the remaining amount of time that it
survives is the same as the original life time
distribution.

 The distribution of additional lifetime is exactly same

as the original distribution of lifetime.

 That is, the distribution of remaining lifetime is9

independent of current age.
 EXERCISE 4.4 – PROBLEM 61
 Data collected at Toronto Pearson International Airport
suggests that an exponential distribution with mean
value 2.725 hours is a good model for rainfall duration.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 a. What is the probability that the duration of a
particular rainfall event at this location is at least 2
hours? At most 3 hours? Between 2 and 3 hours?

 b. What is the probability that rainfall duration exceeds

the mean value by more than 2 standard deviations?
What is the probability that it is less than the mean
value by more than one standard deviation? 10
 SOLUTION:
 λ = 1/(2.725) = 0.3670. µ = 2.725 and σ = 2.725.

 a. What is the probability that the duration of a

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
particular rainfall event at this location is at least 2
hours? At most 3 hours? Between 2 and 3 hours?

Answer:

P ( X  2)  1  F (2)  1  1  e 2    e 2(0.367)  0.48.

P ( X  3)  F (3)  1  e 3(0.367)  0.6675.
P (2  X  3)  F (3)  F (2)  0.6675  0.52  0.1475. 11
 SOLUTION:
 b. What is the probability that rainfall duration exceeds
the mean value by more than 2 standard deviations?
What is the probability that it is less than the mean

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
value by more than one standard deviation?

Answer:

P ( X    2 )  P ( X  8.175)
 1  F (8.175)  1  1  e  (8.175)(0.367)   0.05.
P ( X     )  P ( X  5.45)
 F (5.45)  1  e  (5.45)(0.367)   0.8647. 12
EXERCISE 4.4 – PROBLEM 69
 A consumer is trying to decide between two long-
distance calling plans. The first one charges a flat rate
of 10$ per minute, whereas the second charges a flat
rate 99$ for calls up to 20 minutes in duration and

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
then 10$ for each additional minute exceeding 20
(assume that calls lasting a non integer number of
minutes are charged proportionately to a whole-
minute’s charge). Suppose the consumer’s distribution
of call duration is exponential with parameter λ.

 Which plan is better if expected call duration is 10

minutes? 15 minutes?
13
 THE GAMMA FUNCTION
 For n > 0, the gamma function (n) is defined by

(n)   e  x x n 1 dx.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
0

 The important properties of the gamma function are as

follows.

1. For any n  1, ( n)  ( n  1)( n  1).

2. (1)  1.

 2 
3.  1 .
14
 THE GAMMA FUNCTION
 For any real number n, the gamma function is defined
by

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 (n  1)(n  1), for n  R  ,

 (n  1)!, for n  N ,
(n)   (n  1)
 n , for n  R  \ 1, 2,... ,

 Undefined, for n  0, 1, 2,....

15
THE GAMMA DISTRIBUTION
 A continuous random variable X is said to have a
Gamma distribution if and only if its pdf is given by
 1  1  x 
  x e , x  0,
f ( x;  ,  )    ( )

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
0,
 otherwise.
 Here α and β are known as the parameters of Gamma
distribution.

 The Exponential distribution results from Gamma

distribution by taking α = 1 and β = 1/λ.
16
THE STANDARD GAMMA DISTRIBUTION
 A continuous random variable X is said to have a
Standard Gamma distribution if and only if its pdf is
given by
 1  1  x

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 x e , x  0,
f ( x;  )   ( )
0,
 otherwise.
 Here α is known as the parameter of Standard
Gamma distribution.

 The Standard Gamma distribution results from

Gamma distribution by taking β = 1. 17
GAMMA DENSITY CURVES
For the standard pdf, when α ≤ 1, f(x; α) is strictly
decreasing as x increases from 0.

When α > 1, f(x; α) rises from 0 at x = 0 to a

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
maximum and then decreases. The parameter β is
called the scale parameter.

18
 PROPERTIES
 If X has the Gamma distribution, then

  E ( X )   .

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 2  V ( X )   2 .

1
M X (t )  
,  t  1.
(1   t )

19
 THE INCOMPLETE GAMMA FUNCTION
 The cdf of Standard Gamma distribution is given by

y 1e  y
x
F ( x;  )  P ( X  x)   dy,
( )

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
0

is called the incomplete gamma function.

 Let X have a Gamma distribution with parameters α and

β. Then, for any x > 0, the cdf of X is given by

y 1e  y  x 
x
F ( x;  ,  )  P( X  x)   
dy  F  ;   .
0
 ( )  
20
 EXERCISE 4.4 – PROBLEM 66
 Suppose the time spent by a randomly selected student
who uses a terminal connected to a local time-sharing
computer facility has a gamma distribution with mean

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
20 min and variance 80 min.

 a. What are the values of α and β?

 b. What is the probability that a student uses the

terminal for at most 24 min?

 c. What is the probability that a student spends

21
between 20 and 40 min using the terminal?
 SOLUTION:
 Mean αβ = 20 and αβ2 = 80.

 a. What are the values of α and β?

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
Answer: α = 5 and β = 4.

 b. What is the probability that a student uses the

terminal for at most 24 min?

Answer:
 24  22
P ( X  24)  F (24;5, 4)  F  ;5   F (6;5)  0.715.
 4 
 SOLUTION:
 c. What is the probability that a student spends
between 20 and 40 min using the terminal?

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
Answer:
P (20  X  40)  F (40;5, 4)  F (20;5, 4)
 F (10;5)  F (5;5)  0.971  0.560  0.411.

23
EXERCISE 4.5 – PROBLEM 67
 Suppose that when a transistor of a certain type is
subjected to an accelerated life test, the lifetime X (in
weeks) has a gamma distribution with mean 24
weeks and standard deviation 12 weeks.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 What is the probability that a transistor will last
between 12 and 24 weeks?
 What is the probability that a transistor will last at
most 24 weeks? Is the median of the lifetime
distribution less than 24? Why or why not?
 What is the 99th percentile of the lifetime
distribution?
 Suppose the test will actually be terminated after t
weeks. What value of t is such that only 0.5% of all
transistors would still be operating at termination? 24
SOLUTION:
 Mean αβ = 24 and αβ2 = 144.

 Then, α = 4 and β = 6.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 What is the probability that a transistor will last
between 12 and 24 weeks?

Answer:
P (12  X  24)  F (24; 4.6)  F (12; 4, 6)
 F (4; 4)  F (2; 4)  0.567  0.143  0.424.
25
SOLUTION:
 What is the probability that a transistor will last at
most 24 weeks? Is the median of the lifetime
distribution less than 24? Why or why not?

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
Answer:

P ( X  24)  F (24; 4.6) = F (4; 4) = 0.567.

Yes. The median of the lifetime distribution less than

24.

26
SOLUTION:
 What is the 99th percentile of the lifetime
distribution?

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
Answer: Let A be the 99th percentile of the lifetime
distribution.

Then,
A 
F ( A; 4.6) = F  ; 4  = 0.99
6 
A
  10
6 27
 A  60.
SOLUTION:
 Suppose the test will actually be terminated after t
weeks. What value of t is such that only 0.5% of all
transistors would still be operating at termination?

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
Answer: Since X is the lifetime of the transistor, we
have
P ( X  t )  0.005
 1  F (t ; 4.6) = 0.005
t 
 F  ; 4  = 0.995
6 
t
  11 28
6
 A  66.
THE CHI - SQUARED DISTRIBUTION
 Let υ be a positive integer. A continuous random
variable X with parameter υ is said to have a Chi -
squared (𝜒 2 ) distribution if and only if its pdf is given
by

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 1  2  1  x 2
 2 x e , x  0,
f ( x; )   2 ( 2)
0,
 otherwise.
 The parameter υ of Chi - squared distribution is called
the degrees of freedom (df) of X.

 The Chi - squared distribution results from Gamma

29
distribution by taking α = υ/2 and β = 2.
 PROPERTIES
 If X has the Chi - squared distribution, then

  E ( X )  .

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 2  V ( X )  2.

1
M X (t )  
, 2t  1.
(1  2t ) 2

30
THE LOGNORMAL DISTRIBUTION
 A continuous random variable X is said to have a
Lognormal distribution if the random variable
Y = ln(X) has a normal distribution.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 A continuous random variable X is said to have
Lognormal distribution if and only if its pdf is given
by
 1  ln X   
2
2 2
 e , x  0,
f ( x;  ,  )    2 x
0,
 otherwise.
 Here μ and σ are known as the parameters of
31
Lognormal distribution.
LOGNORMAL DENSITY CURVES
Although a normal curve is symmetric, a
lognormal curve has a positive skew.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
32
Dr Jagan Mohan Jonnalagadda, BITS
Pilani, Hyderabad Campus.
33
 PROPERTIES
 If X has the Lognormal distribution, then
2

E( X )  e 2
.

  e 

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
2   2 2
V (X )  e 1 .

 Because ln(X) has a normal distribution, the cdf of X

can be expressed in terms of the cdf Φ(z) of a standard
normal random variable Z.

F ( x;  ,  )  P( X  x)  P(ln X  ln x)
 ln x     ln x   
 PZ     .
      34
EXERCISE 4.5 – PROBLEM 81
 A theoretical justification based on a certain material
failure mechanism underlies the assumption that
ductile strength X of a material has a lognormal
distribution. Suppose the parameters are μ = 5 and σ
= 0.1.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 Compute E(X) and V(X).
 Compute P(X > 125).
 Compute P(110 ≤ X ≤ 125).
 What is the value of median ductile strength?
 If ten different samples of an alloy steel of this type
were subjected to a strength test, how many would
you expect to have strength of at least 125?
 If the smallest 5% of strength values were
unacceptable, what would the minimum acceptable35
strength be?
 SOLUTION:
 The ductile strength X of a material has a lognormal
distribution. The parameters are μ = 5 and σ = 0.1.

 Compute E(X) and V(X).

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
2

E( X )  e 2
 e5.005 .


V (X )  e 2   2
 2
e   
 1  e10.01 e0.01  1 .

 Compute P(X > 125).

P ( X  125)  1  P ( X  125)  1  P(ln X  ln125)

 ln125  5   ln125 365 
 1 P Z    1  .
 0.1   0.1 
SOLUTION:
 Compute P(110 ≤ X ≤ 125).

 ln125  5   ln110  5 
P (110  X  125)     .
 0.1   0.1 

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 What is the value of median ductile strength?
F ( x;5, 0.1)  0.5
 P ( X  x)  0.5
 ln x  5 
    0.5
 0.1 
ln x  5
 0
0.1 37

 x  e5 .
SOLUTION:
 If ten different samples of an alloy steel of this type
were subjected to a strength test, how many would
you expect to have strength of at least 125?

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
Answer: Let Y be the number of samples of an alloy
strength of this type have strength of at least 125.

Then, Y ~ Bin(n, p) with n = 10 and p = 0.9573.

E(Y) = np = 9.573

38
SOLUTION:
 If the smallest 5% of strength values were
unacceptable, what would the minimum acceptable
strength be?

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
F (t ;5, 0.1)  0.05
 P ( X  t )  0.05
 ln t  5 
    0.05
 0.1 
ln t  5
  1.645
0.1
 t  e 4.8355 . 39
BETA FUNCTION

 The Beta function is defined by

1 1

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
1  x  1  x 
 1  1
B ( ,  )   x  1
dx   x  1
dx,
0 0

for any α, β > 0.

 The relationship between gamma function and beta

function is given by

( )(  )
B ( ,  )  .
(   ) 40
 THE BETA DISTRIBUTION
 A continuous random variable X is said to have a Beta
distribution if and only if its pdf is given by
 1 (   )  x  A  1  B  x   1
     , A  x  B,

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
f ( x;  ,  )   B  A ( )(  )  B  A   B  A 
0,
 otherwise.

 Here α and β (both positive) are known as the parameters

of Beta distribution.

 The case A = 0 and B = 1 gives the standard beta

distribution.
41
 PROPERTIES
 If X has the Beta distribution, then

  

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
  E( X )  A  B  A  .
   

( B  A )2

 V(X) 
2
.
(   ) (    1)
2

42
 Dr Jagan Mohan Jonnalagadda, BITS
Pilani, Hyderabad Campus.
43
BETA DENSITY CURVES
EXERCISE 4.5 – PROBLEM 84
 Suppose the proportion X of surface area in a
randomly selected quadrat that is covered by a
certain plant has a standard beta distribution with α
= 5 and β = 2.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 a. Compute E(X) and V(X).

 b. Compute P(X ≤ 0.2).

 c. Compute P(0.2 ≤ X ≤ 0.4).

 d. What is the expected proportion of the sampling

region not covered by the plant?
44
SOLUTION:
 X, the proportion of the surface area in a randomly
selected quadrat that is covered by a certain plant
has a standard beta distribution with α = 5 and β = 2.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 a. Compute E(X) and V(X).

   5
  E( X )    .
    7
( B  A )2
 10
 V(X) 
2
 .
(   ) (    1) 392
2

45
SOLUTION:
 b. Compute P(X ≤ 0.2).
0.2 0.2
P  X  0.2    x 1  x  dx   x 1  x  dx.
 1  1 4

0 0

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
 c. Compute P(0.2 ≤ X ≤ 0.4).
0.4
P  0.2  X  0.4    x 1  x 
4
dx.
0.2

 d. What is the expected proportion of the sampling

region not covered by the plant?

5 2
E (1  X )  1   . 46
7 7
IMPORTANT INTEGRALS - I

 
 ( n)   e x
x n 1
dx  2  e  z2
z 2 n 1 dz.

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS
0 0

(1)  1  2  e  z2
zdz.
0

1
      2  e dz.
 z2

2 0

47
IMPORTANT INTEGRALS - II

 
 ( n)   e x
x n 1
dx  2 1 n
e
 z2 2
z 2 n 1 dz.
0 0

Pilani, Hyderabad Campus.

Dr Jagan Mohan Jonnalagadda, BITS

(1)  1   e  z2 2
zdz.
0

1
     2e  z2 2
dz.
2 0

3  1

 z2 2 2
    e z dz.
2 2 20
48
Dr Jagan Mohan Jonnalagadda, BITS
Pilani, Hyderabad Campus.
49