Gamma and Beta

UNIT 16 GAMMA AND BETA Distributions

DISTRIBUTIONS
Structure
16.1 Introduction
Objectives

16.2 Beta and Gamma Functions

16.3 Gamma Distribution
16.4 Beta Distribution of First Kind
16.5 Beta Distribution of Second Kind
16.6 Summary
16.7 Solutions/Answers

16.1 INTRODUCTION
In Unit 15, you have studied continuous uniform and exponential
distributions. Here, we will discuss gamma and beta distributions. Gamma
distribution reduces to exponential distribution and beta distribution reduces
to uniform distribution for special cases. Gamma distribution is a
generalization of exponential distribution in the same sense as the negative
binomial distribution is a generalization of geometric distribution. In a sense,
the geometric distribution and negative binomial distribution are the discrete
analogs of the exponential and gamma distributions, respectively. The present
unit discusses the gamma and beta distributions which are defined with the
help of special functions known as gamma and beta functions, respectively.
So, before defining these distributions, we first define gamma and beta
functions in Sec. 16.2 of this unit. Then gamma distribution and beta
distribution of first kind followed by beta distribution of second kind are
discussed in Secs. 16.3 to 16.5.
Objectives
After studing this unit, you would be able to:
 define beta and gamma functions;
 define gamma and beta distributions;
 discuss various properties of these distributions;
 identify the situations where these distributions can be employed; and
 solve various practical problems related to these distributions.

16.2 BETA AND GAMMA FUNCTIONS

In this section, some special functions i.e. beta and gamma functions are
defined with their properties and the relation between them. These will be
helpful in defining beta and gamma distributions to be defined in the
subsequent sections.

71
 Continuous Probability
Distributions Beta Function
1
m 1 n 1
Definition: If m > 0, n > 0, the integral  x 1  x 
0
dx is called a beta

function and is denoted by β(m, n) e.g.

1 1 3
2 1 31 3 
i)  x 1  x  dx   x 2 1  x  dx    ,3 
0 0 2 
1
2 1  3 
or  x 1  x  dx     1, 2  1    , 3 
0 2  2 
1 1 1
   1 1  2 2
ii) x 3
1  x  3 dx      1,   1    , 
0  3 3  3 3

Properties of Beta Function

1. Beta function is symmetric function i.e. β(m, n) = β(n, m)
2. There are some other forms also of Beta function. One of these forms,
which will be helpful in defining beta distribution of second kind, is

x m 1
  m, n    mn
dx
0 1  x 
  p,q  1   p  1, q 
3. (i) 
q p
(ii) β(p, q) = β(p + q, q)  β(p, q + 1)

On the basis of the above discussion, you can try the following exercise.
E1) Express the following as a beta function:
1 1 1

i) x 3
1  x  2 dx
0

1
2 5
ii)  x 1  x 
0
dx


x2
iii)  1  x  5
dx
0

1
 
2
x
iv)  1  x  2
dx
0

72
Gamma Function Gamma and Beta
Distributions
Though we have defined Gamma function in Unit 13, yet we are again
defining it with more properties, examples and exercises to make you clearly
understand this special function.

n 1  x
Definition: If n > 0, the integral x e dx is called a gamma function and is
0

denoted by  n 

e.g.

2 x
(i) x e dx   2  1   3
0


x 1  3
(ii)  xe dx    1    
0 2  2
Some Important Results on Gamma Function

1. If n > 1,  n    n  1  n  1
2. If n is a positive integer, n   n  1 !

1
3.    
2
Relationship between Beta and Gamma Functions

If m > 0, n > 0, then   m, n  

 m n 
m  n
You can now try the following exercise.
E2) Evaluate:
 5
x
(i)  e x 2 dx
0


10
(ii)  1  x 
0
dx

 1

x
(iii)  x 2 e dx
0

16.3 GAMMA DISTRIBUTION

Gamma distribution is a generalisation of exponential distribution. Both the
distributions are good models for waiting times. For exponential distribution,
the length of time interval between successive happenings is considered i.e.
the time is considered till one happening occurs whereas for gamma
distribution, the length of time between 0 and the instant when rth happening
73
 Continuous Probability
Distributions
occurs is considered. So, if r = 1, then the situation becomes the exponential
situation. Let us now define gamma distribution:
Definition: A random variable X is said to follow gamma distribution with
parameters r > 0 and  > 0 if its probability density function is given by
  r e x x r 1
 , x0
f (x)   r 

0, elsewhere

Remark 1:
(i) It can be verified that


 f  x  dx  1
0

Verification:
 
 r ex x r 1
  x dx  
0 0 r
dx

r 1

ex  x 
 dx
0 r
Putting  x = y   dx  dy
Also, when x  0, y  0 and when x  , y  

1 y
 e y r 1dx
r 0

1
 r [Using gamma function defined in Sec. 16.2]
r
=1
(ii) If X is a gamma variate with two parameters r > 0 and  > 0, it is expressed
as X  γ(, r).
(iii) If we put r = 1, we have

e x x 0
f (x)  ,x  0
1
 e x , x  0
which is probability density function of exponential distribution.

Hence, exponential distribution is a particular case of gamma

distribution.

(iv) If we put  = 1, we have

e x .x r 1
f (x)  , x  0, r  0
r
74
It is known as gamma distribution with single parameter r. This form of the Gamma and Beta
gamma distribution is also widely used. If X follows gamma distribution with Distributions

single parameter r > 0, it is expressed as X   (r).

Mean and Variance of Gamma Distribution
If X has a gamma distribution with parameters r > 0 and  > 0, then its
r r
Mean = , Variance = 2 .
 
If X has a gamma distribution with single parameter r > 0, then its
Mean = Variance = r.
Additive Property of Gamma Distribution
1. If X1, X2, …, Xk are independent gamma variates with parameters
 , r1  ,  , r2  ,...,  , rk  respectively, then X1 + X2 +…+ Xk is also a
gamma variate with parameter (, r1 + r2 +…+rk).
2. If X1, X2,...,Xk are independent gamma variates with single parameters r1,
r2,…, rk respectively, then X1 + X2 + …, + Xk is also a gamma variate with
parameter r1 + r2 + … + rk.
Example 1: Suppose that on an average 1 customer per minute arrive at a
shop. What is the probability that the shopkeeper will wait more than 5
minutes before
(i) both of the first two customers arrive, and
(ii) the first customer arrive?
Assume that waiting times follows gamma distribution.
Solution:
i) Let X denotes the waiting time in minutes until the second customer
arrives, then X has gamma distribution with r = 2 (as the waiting time is to
be considered up to 2nd customer)
 = 1 customer per minute.
 
 r ex x r 1
 P  X  5   f (x)dx   dx
5 5 r
2

1 e x x 21 
e  x x1 
 dx   dx   x1e  x dx
5  2 5
1 5

  e  x   e  x 
 x    1 dx  [Integrating by parts]
  1 5 5 1 

 
 ex 
  0  5e5    e x dx  5e5   
5  1  5
 5e 5   0  e5 

= 6 e 5
75
 Continuous Probability
Distributions
= 6  0.0070 [See the table given at the end of Unit 10]
= 0.042
ii) In this case r = 1,  = 1 and hence

 r ex .x r 1
P  X  5   dx
5 r
  
(1)1 e  x x 0  e x 
 dx   e x dx     0  e5  0.0070
5 (1) 5  1  5
Alternatively,
As r = 1, so it is a case of exponential distribution for which
f  x   ex , x  0

  
 e x 
 P  X  5    e x x
dx   1e dx     0  e 5  0.0070
5 5  1  5
Here is an exercise for you.
E3) Telephone calls arrive at a switchboard at an average rate of 2 per minute.
Let X denotes the waiting time in minutes until the 4th call arrives and
follows gamma distribution. Write the probability density function of X.
Also find its mean and variance.

Let us now discuss the beta distributions in the next two sections:

16.4 BETA DISTRIBUTION OF FIRST KIND

You have studied in Sec. 16.3 that beta function is related to gamma function
in the following manner:

  m, n  
 m n 
m  n
Now, we are in a position to define beta distribution which is defined with the
help of beta function. There are two kinds of beta distribution  beta
distribution of first kind and beta distribution of second kind. Beta distribution
of second kind is defined in next section of the unit whereas beta distribution
of first kind is defined as follows:
Definition: A random variable X is said to follow beta distribution of first
kind with parameters m > 0 and n > 0, if its probability density function is
given by

 1 m 1 n 1
  m, n x 1  x  , 0  x  1
f (x)    
0, otherwise

The random variable X is known as beta variate of first kind and can be
expressed as X  1(m, n)
76
Remark 5: If m = 1 and n = 1, then the beta distribution reduces to Gamma and Beta
Distributions
1 11
f x  x11 1  x  , 0  x  1
 1,1
0
x 0 1  x 
 , 0  x 1
 1,1

1
 ,0  x 1
 1,1

11 0 0
But  1,1  
2 1

Therefore, f (x) 
11  1
1
 f (x)  1, 0  x  1
1
 ,0  x 1
1 0
which is uniform distribution on (0, 1).
1
[p.d.f. of uniform distribution on (a, b) is f (x)  , a  x  b]
ba
So, continuous uniform distribution is a particular case of beta
distribution.
Mean and variance of Beta Distribution of First Kind
Mean and Variance of this distribution are given as
m
Mean =
mn
mn
Variance = 2
 m  n   m  n  1
Example 4: Determine the constant C such that the function
6
f (x)  Cx 3 1  x  , 0  x  1 is a beta distribution of first kind. Also, find its
mean and variance.
Solution: As f  x  is a beta distribution of first kind.
1
 f x  1
0

1
3 6
  Cx 1  x 
0
dx  1

1
6
 C  x 3 1  x  dx  1
0

77
 Continuous Probability
Distributions  C   3  1, 6  1  1 [By definition of Beta distribution of first kind]

1
 C
  4, 7 

47  m n 
    m, n   
47   m  n  
11 10
 
4 7 3 6

10  9  8  7  6
  840
3 2  6
6
Thus, f  x   840x 3 1  x 
7 1
= 840x 4 1 1  x 
7 1
x 41 1  x 

  4, 7 

1
[  840 just obtained above in this example]
  4, 7 

 m = 4, n = 7
m 4 4
 Mean =   ,
m  n 4  7 11
mn
and Variance = 2
 m  n   m  n  1
47
 2
 4  7   4  7  1
28 7 7
  
12112 121 3 363
Now, you can try the following exercises.
E4) Using beta function, prove that
1
2 3
 60x 1  x 
0
dx  1

E5) Determine the constant k such that the function

1 1

f (x)  kx 2 1  x  2 ,0  x  1, is a beta distribution of first kind. Also
find its mean and variance.

78
 Gamma and Beta
16.5 BETA DISTRIBUTION OF SECOND KIND Distributions
Let us now define beta distribution of second kind.
Definition: A random variable X is said to follow beta distribution of second
kind with parameters m > 0, n > 0 if its probability density function is given
by
 x m 1
 mn
, 0x
f  x      m, n 1  x 

0, elsewhere

x m 1
Remark 6: It can be verified that    m, n 1  x  mn
dx  1
0

Verification:
 
x m 1 1 x m 1
   m, n 1  x  dx  dx
0
m n
  m, n  0 1  x  m  n

  x m-1 
  mn
dx is another form 
 0 1+x  
1 of beta function. 
   m, n 
  m, n   
(see Sec. 16.2 of this Unit) 
 
 
=1

Remark 7: If X is a beta variate of second kind with parameters m > 0, n > 0,

then it is expressed as X  2(m, n)
Mean and Variance of beta Distribution of second kind
m
Mean = , n  1;
n 1
m  m  n  1
Variance = 2
,n  2
 n  1  n  2 
Example 5: Determine the constant k such that the function
kx 3
f x  7
, 0  x  ,
1  x 
is the p.d.f of beta distribution of second kind. Also find its mean and
variance.
Solution: As f  x  is a beta distribution of second kind,


  f  x  dx  1
0

79
 Continuous Probability 
Distributions kx 3
  1  x  dx  1
0
7


x 4 1
 k 4 3
dx  1
0 1  x 
 k  4, 3  1

1 7 6 6 5 4
 k     60
  4,3 4 3 3 2 2

Here m = 4, n = 3
m 4 4
 Mean =   2
n 1 3  1 2
m  m  n  1 4(4  3  1) 46
Variance = 2
 2
 6
 n  1  n  2  (3  1)  3  2  4 1

Now, you can try the following exercises.

E6) Using beta function, prove that

x3 64
 13
dx 
15015
0 1  x  2

E7) Obtain mean and variance for the beta distribution whose density is given
by
60x 2
f x  7
,0  x  
1  x 

16.6 SUMMARY
The following main points have been covered in this unit:
1) A random variable X is said to follow gamma distribution with
parameters r > 0 and  > 0 if its probability density function is given by
  r ex x r 1
 , x0
f x   r

0, elsewhere

2) Gamma distribution of random variable X with single parameter r > 0 is

e  x x r 1
defined as f  x   , x  0, r  0
r
r
3) For gamma distribution with two parameters λ and r, Mean = and

r
Variance = .
2
80
4) A random variable X is said to follow beta distribution of first kind with Gamma and Beta
parameters m > 0 and n > 0, if its probability density function is given by Distributions

 1 m 1 n 1
  m, n x 1  x  , 0  x  1
f x    
0, otherwise

m mn
Its mean and variance are and 2
, respectively.
mn  m  n   m  n  1
5) A random variable X is said to follow beta distribution of second kind
with parameters m > 0, n > 0 if its probability density function is given by:
 x m 1
 mn
,0  x  
f  x      m, n 1  x 

0, elsewhere

m m  m  n  1
Its Mean and Variance are , n  1; and 2
,n  2
n 1  n  1  n  2 
respectively.
6) Exponential distribution is a particular case of gamma distribution and
continuous uniform distribution is a particular case of beta distribution.

16.7 SOLUTIONS/ANSWERS
1 1 1
1 1  2 3
1  x  2 dx  B  

3
E1) (i) x  1,  1  B  , 
0  3 2  3 2
1 1
2 11 5 6 1
(ii)  x 1  x  dx   x 1  x 
0 0
dx

is not a beta function, since m =  1 < 0, but m and n both should be

positive.
 
x2 x 31
(iii)  5
dx   3 2
dx = β(3, 2)
0 1  x  0 1  x 
[ m = 3, n = 2(see Property 2 of Beta function Sec. 16.2)]
1 1
   1
2
x x2 1 3
(iv)  1  x  2
dx   1 3
dx    , 
0 0 1  x 

2 2 2 2


5 
E2)  e x .x 5/ 2 dx    1 
0 2 

7
  
2

81
 Continuous Probability
Distributions  5  5  5   3   3   5  3   1   1 
                 
 2  2  2   2   2   2  2   2   2 
 Result 1on gamma 
function (See Sec. 16.2) 
 
 5  3  1 
       Result 3 on gamma function 
 2  2  2 

 15 
  
8

1 10
(ii)  x 1  x 
0
dx = β(1 + 1, 10 + 1)

= β(2,11)

2 11 see relation between 

=  beta and gamma function 
13  

=
1!10 ! [Result 2 on gamma function]
12 !

=
10  ! =
1

1
12 1110 ! 12 11 132
 1

x  1  1
(iii) 0 2 e dx    2  1   2   
x

E3) Here  = 2, r = 4.
 r ex .x r 1
 f (x)  ,x  0
r 
24.e2x .x 3
 ,x  0
4
16e2x .x 3
= ,x  0
3

8
= x 3e 2x , x  0
3
r 4
Mean =   2,
 2
r 4
Variance = 2
 2 1
 2
1 1
3 4 1
E4)  60x 2 1  x  dx  60  x 31 1  x  dx  60   3, 4 
0 0

82
 34 2 3 60  2  3  2 Gamma and Beta
= 60 = 60   1 Distributions
7 6 6  5  4  3 2
1 1 1

E5)  kx 2
1  x  2 dx  1
0

 1 1 
 k    1,  1  1
 2 2 

1 2 1 2 2
 k    
1 3   
 ,  1 3 1
 
1
 
2 2 2 2 2 2
Now, as the given p.d.f. of beta distribution of first kind is
2  12 1
f (x)  x 1  x  2 , 0  x  1

1 3
1 1
x2 1  x  2
 , 0  x 1
1 3
 , 
2 2
1 3
m  , n 
2 2
1
m 1
and hence mean =  2 
mn 1  3 4
2 2
mn
Variance = 2
 m  n   m  n  1
 1  3  3
   3 1
2
   2 4
 2
 2
 
 1 3   1 3   2   3 4  4  3 16
      1
2 2 2 2 
 
x3 x 4 1
E6)  1  x  13/ 2
dx   5
dx
4
0 0 1  x  2

5 5
4 3
 5  2  2
=   4,  
 2  5 13
4  2
 2

5
6
2 6  32 64
=  
13 11 9 7 5 5 13 11 9  7  5 15015
. . . .
2 2 2 2 2 2
83
 Continuous Probability
Distributions 60x 2
E7) f  x   7
,0  x  
1  x 
60x 31
 3 4
,0  x  
1  x 
x 31  34 23 1
 3 4
, 0x    3, 4     
  3, 4 1  x   6 6 60 

m =3, n = 4
m 3
Hence, mean =  1
n 1 4  1
m  m  n  1 3  3  4  1 3 6
Variance = 2
= 2
 = 1.
 n  1  n  2   4  1  4  2  9 2

84