3.2.

4 The Beta Distribution

Consider the beta function defined by

1
B( ,  )   t  1 (1  t )  1 dt …………………………………………………………………..(1)
0

On dividing both sides of (1) by B( ,  ) we get

1 1
1
B( ,  ) 0
t  1 (1  t )  1dt ………………………………………………………………….(2)

Because an integration of a pdf over the entire sample space is always unit, we

conclude that the integrand of (2) is a valid pdf.

 x 1 1  x   1
 , 0<x<1;  ,   0 
Therefore f ( x)     ,   is a pdf and is called a beta

0, elsewhere

density function. Hence any random variable X which has this probability

distribution is said to be Beta distributed.

A relationship between beta and gamma functions is given by

( )(  )
B( ,  )  and in view of this relationship, the beta function can also be
(   )

defined by the density function

 (   )  1
x 1  x  , 0<x<1;  ,   0 
 1

f ( x)   ( )(  )
0,
 elsewhere

1
Some distributional quantiles

 
The mean and variance are   and  2 
          1
2

Proof

1
E ( X r )   x r f ( x)dx
0

x 1 1  x 
 1
1
 x r
dx
0   ,  
1 1

B ( ,  ) 0
x r  1 (1  x)  1 dx

1
 B (  r ,  )
B ( ,  )
(   ) (  r )(  )
 
( )(  ) (    r )
(   )(  r )(  )

( )(  )(    r )

Therefore

(   )(  1)(  )

E( X ) 
( )(  )(    1)
(   )( )(  )

( )(  )(   )(   )


 

2
 (   )(  2)(  )
E( X 2 ) 
( )(  )(    2)
(   )(  1)(  1)(  )

( )(  )(    1)(    1)
(   ) (  1)( )(  )

( )(  )(   )(    1)(   )
 (  1)

(   )(    1)

Hence

Var ( X )  E ( X 2 )   E ( X ) 
2

2
 (  1)   
  
(   )(    1)     
 (  1)(   )   2 (    1)

(   ) 2 (    1)
( 2   )(   )  ( 3   2    2 )

(   ) 2 (    1)
 3   2    2    ( 3   2    2 )

(   ) 2 (    1)


(   ) (    1)
2

3.2.5 The Normal Distribution

One of the most important examples of a continuous probability distribution is

the normal distribution, normal curve, or Gaussian distribution. The density

function for this distribution is given by:

 1 1 
2
x

 e 2  , - <x<
f ( x)    2
0,
 elsewhere

Where   mean,   standard deviation,   3.14159 and e  2.71828

3
 1 
2
x
1
The total area bounded by the curve Y  e 2  and the x  axis is 1; hence
 2

the area under the curve between two coordinates x  a and x  b , where a  b ,

represents the probability that x lies between a and b . This probability is

denoted by P(a  X  b) .

Construction of the normal distribution

Consider the integral

y2
 
I  e 2
dy


This integral exists because the integrand is a positive continuous function

which is bounded by an integrable function; that is

y2
 
0e 2
 e( | y|1) ,   y   and 

e( | y|1) dx  2e

To evaluate the integral I , we note that I  0 and that I 2 may be written

 y2  z2 

   2 
I   e  
dy dz
 

This iterated integral can be evaluated by changing to polar coordinates.

If we set y  r cos  and z  r sin  , we have

4
 r2
2  
I 
2
 e 2
r dr d
0 0
2
  d
0

 2

2
 1  y2
Accordingly I  2 and  
2
e dy  1 .

If we introduce a new variable of integration, say x , by writing

x
y ,   0 , the preceding integral becomes


 ( x   )2 
  
1  2 2 
  2 
e dx  1 .

Since   0 , this implies that

( x   )2
1 
f ( x)  e 2 2
,   x   satisfies the condition of being a pdf of a
 2

continuous type random variable.

A random variable of a continuous type that has a pdf of the form f ( x ) is said

to have a normal distribution, and any f ( x ) of this form is called a normal pdf.

The distribution function for a random variable X which is normally distributed

with mean μ and variance σ2 is given by:

1  u  
2

1 x   


2  
F ( x)  P( X  x)  e du ……………………………………………………….(1)
 2 

5
If X has a distribution function given by equation (1) above, we say that the

random variable X is normally distributed with mean μ and variance σ2. If we

x
let Z be the standardized variable corresponding to X i.e. we let z  , then


the mean or the expected value of Z is 0 and variance is 1. In such cases, the

density function of Z can be obtained from the density function of X by formally

z2
1 2
plugging in μ= 0 and σ= 1, yielding f ( z )  e .
2

This is often referred to as the standard normal density function. The

corresponding distribution function is given by

2
z u2 z u
1  1 1 

e 2 0
F ( z )  p(Z  z )  2
du =  e 2
du
2 
2

We sometimes call the value z of the standardized variable Z the standard score.

z2
1 2
A graph of the density function f ( z )  e , sometimes called the standard
2

normal curve, is as shown below:

In this graph, the area with 1 standard deviation of the mean is indicated as

68.26% or 0.6826. This means that P (1  Z  1) = 0.6826. A table giving areas

6
under this curve bounded on the ordinate z  0 and any positive value of z is

provided in appendices in most of the statistical books. From this table the areas

between any two ordinates can be found by using symmetry of the curve about

z  0.

Example 1

Find the area under the standard normal curve

a) Between z=0 and z=1.2

b) Between z=-.68 and z=0

c) Between z=-.46 and z=2.21

d) Between z=.81 and z=1.94

e) To the right of z=1.28

Solution

a) .3849

b) .2518

c) .6636

d) .1828

e) .8997

Example 2

If ‘area’ refers to that under the standard normal curve, find the value or values

of z such that

7
 a) Area between 0 and z is .3770

b) Area to the left of z is .8621

Solution

a) z = ±1.16

b) z = 1.09

Example 3

The mean weight of 500 male students at Kisii University is 151 lbs and the

standard deviation is 15 lbs. Assuming that the weights recorded to the nearest

pound and are normally distributed, find how many students weight

a) between 120 and 155 lbs

b) more than 185 lbs

Solution

a) 300

b) 5

Some distributional quantiles

The moment generating function of a random variable X ~ N (  ,  2 ) is derived as

the following.

8
 1  x 
2

 1    
M (t )  E (e tx )   e tx f ( x)dx   e tx e 2  
dx
 
 2
1  x 
2

 1   
  e tx e t e t e 2  
dx

 2
1  x 
2

e t    


t ( x ) 2  
 e e dx
 2 

e t  
1
 x  2 t  x 
2 2


 2 

e 2 2
dx

On completing the square of the exponent we obtain

e t  
1
 x  2 t  x  t  t 
2 2 4 2 4 2

M (t ) 
 2 

e 2 2
dx
1
 2t
e t e 2  
1
 x   t 
2 2


 2 

e 2 2
dx

 2 x (   2t 
1 1 2
t   2 t  1
e 2
  2 e 2 dx
1
t   2 t
e 2

 2 x (   2t 
1 2
1 
because  e 2 dx  1 .

 2

Note

We can use the above result in showing that the mean and variance are

respectively  and  2 .