3.2.3.

THE BETA DISTRIBUTION

The uniform density

f (x)=¿ {1 , for 0<x<1¿¿¿¿
is a special case of the beta
distribution, which is defined in the following way:

Definition 3.2.5.
A random variable X has a beta distribution and it is referred to as a beta random
variable if and only if its probability density function is given by

Γ(α+β) α−1
f (x )=¿
{ Γ(α) Γ(β )
x (1−x)β−1 , for 0<x<1 ¿ ¿¿¿

where α , β >0 are the parameters of the distribution.

In recent years, the beta distribution has found important applications in Bayesian
inference, where the parameters are looked upon as random variables, and there is a
need for a fairly “flexible” probability density for the parameter p of the binomial
distribution, which takes on nonzero values only on the interval from 0 to 1. By
“flexible” we mean that the probability can take on a great variety of different shapes.
The use of the beta distribution will be discussed later on.

We shall not prove here that the total area under the curve of the beta distribution, like
that of any probability density, is equal to 1, but in the proof of the theorem that
follows, we shall make use of the fact that

1
Γ (α+ β ) α−1
∫ x (1−x )β −1 dx=1
x=0 Γ (α ) Γ ( β )

and hence that

1
Γ (α )⋅Γ ( β )
∫ x α −1(1−x )β−1 dx= Γ (α+β )
x =0

This integral defines the beta function, whose values are denoted B ( α , β ) ; in other
Γ (α )⋅Γ ( β )
B ( α , β )=
words, Γ (α+ β ) .
Detailed discussion of the beta function may be found in any textbook on advanced
calculus.

Page 1 of 8
It is quite straight forward to evaluate the moments of a Beta random variable.

1
kΓ (α +β )
E( X )= ∫ x k +α−1(1−x )β−1 dx
Γ (α ) Γ ( β ) x=0

Γ (α+β ) Γ (α+k ) Γ ( β ) Γ (α +β ) Γ (α+k )

= =
Γ (α ) Γ ( β ) Γ (α+β +k ) Γ (α ) Γ (α+ β+k )

Theorem 3.2.1

The mean of the Beta distribution is

α
E( X )=
α+ β and the variance is

αβ
Var ( X )= 2
(α+ β ) (α+ β+1 )

Proof

From

Γ (α+β ) Γ (α+k )
E( X k )=
Γ( α) Γ ( α+β +k )

If k = 1, we have

Γ ( α+β ) Γ (α +1)
E( X )=
Γ ( α) Γ (α +β+1)
( α+β−1 ) ! α !
=
( α−1 ) ! ( α+β ) !
( α+β−1 ) ! α⋅(α−1 )!
=
( α−1 ) ! ( α +β )⋅( α+ β−1 ) !
α
=
α+ β

α
⇒ Mean = E( x )=
α+β

Page 2 of 8
 Γ (α + β ) Γ (α +2 )
E( X 2 )=
Γ (α ) Γ (α + β +2 )
( α + β−1 ) ! ( α +1) !
=
( α −1 ) ! ( α + β +1 ) !
( α+ β−1 ) ! ( α + 1)⋅α⋅( α−1 )!
=
( α−1 ) ! ( α + β+ 1 )⋅( α+ β )⋅( α + β−1 ) !
α ( α+1 )
=
( α + β +1) ( α + β )

2
⇒ Var ( X ) = E( x )−[ E( x ) ]
2
α( α +1) α
= − [ ]
( α+ β +1 ) ( α + β ) α + β
α( α +1) α2
= −
( α+ β +1 ) ( α + β ) ( α + β )2
α (α +1 )( α+ β )−α 2 (α + β +1 )
=
( α + β+ 1) ( α + β )2
αβ
=
( α+ β +1) ( α + β )2

αβ
i. e . Var ( X )=
( α + β +1) ( α+ β )2

The moment generating function for the Beta distribution is neither simple nor useful.
The Beta density can take on a variety of shapes, including the uniform on (0, 1) with
α =β=1 .
Figure 1 indicates how these shapes change with the values for the parameters.

Page 3 of 8
3.2.4. THE NORMAL DISTRIBUTION

The normal distribution, which we shall study in this section, is in many ways the
cornerstone of modern statistical theory. It was investigated first in the eighteenth
century when scientists observed an astonishing degree of regularity in errors of
measurement. They found that the patterns (distributions) that they observed could be
closely approximated by continuous curves, which they referred to as “normal curves
of errors” and attributed to the laws of chance.

Importance of Normal Distribution

The normal distribution has great significance in statistical work because of the
following reasons:
1. The normal distribution has the remarkable property stated in the so called
central limit theorem, which asserts that certain statistics, most important of
which is the sample mean and sample variance, tends to be normally distributed
as the sample size becomes large.
2. Even if a variable is not normally distributed, it can sometimes be brought to
normal form by simple transformation of variable. For example, if distribution of
X is skewed, the distribution of √ X might come out to be normal.
3. Many of the sampling distributions like student’s t, F, etc., also tend to normal
distribution as the sample size becomes larger.
4. The sampling distribution and tests of hypothesis are based upon the assumption
that samples have been drawn from a normal population with mean μ and
2
variance σ .
5. Normal distribution find large applications in Statistical Quality Control.
6. As n becomes large, the normal distribution serves as a good approximation for
many discrete distributions (such as Binomial, Poisson, etc.)
7. In theoretical statistics, many problems can be solved only under the assumption
of a normal population. In applied work, we often find that methods developed
under the normal probability law yield satisfactory results, even when the
assumption of a normal population is not fully met, despite the fact that the
problem can have a formal solution only if such a premise is hypothesized.
8. The normal distribution has numerous mathematical properties which make it
popular and comparatively easy to manipulate.

Definition 3.2.6
A random variable X has a normal distribution and it is referred to as a normal random
variable if and only if its probability density function is given by

Page 4 of 8
 1 x−μ 2
N ( x; μ, σ 2 ) =f ( x)=¿ { 1
σ √2π
−
e
( ),
2 σ
for −∞<x<∞ ¿ ¿¿¿
………………………..(3.2.3)
where −∞< μ<∞ , σ >0
A random variable having this distribution is said to be normal or normally
distributed. This distribution is very important, because many random variables of
practical interest are normal or approximately normal or can be transformed into
normal random variables in a relatively simple fashion. Furthermore, the normal
distribution is a useful approximation of more complicated distributions. It also appears
in the mathematical proofs of various statistical tests.
In Eq. (3.2.3), μ is the mean and σ is the standard deviation of the distribution.
The curve of f (x) is called the bell-shaped curve. It is symmetric with respect to μ .

The figure above shows f (x) for μ=0 . For μ>0 ( μ<0 ) the curves have the same

shape, but are shifted |μ| units to the right (left).

2
The smaller σ is, the higher is the peak at x = 0 in figure (1) above and the steeper are
the descents on both sides. This agrees with the meaning of variance.

First, though, let us show that the formula of Definition 3.2.6 can serve as a probability
2
density. Since the values of N ( x; μ , σ ) are evidently positive as long as σ >0 , we
must show that the total area under the curve is equal to 1.
Integrating from −∞ to ∞ and making the substitution
x−μ dz 1 1
z= ⇒ = ⇒ dz= dx
σ dx σ σ , we get

Page 5 of 8
 2
1 x− μ
∞ − ( ∞
) dx 1 2 1 2
1 1 ∞ −2 z 2 ∞ −2 z
∫ f (x ) dx=1 ⇒ ∫ σ √2 π 2
e
σ
⇒ ∫
√ 2 π x= −∞
e dz= ∫ e dz
√ 2 π x= 0
x = −∞ x= −∞

Γ ( 12 ) = √ π
Then, since the integral on the right equal √2 √2 according to Equation (3.2.2), it
2 √π
× =1
follows that the total area under the curve is equal to √2 π √ 2

The Moment generating function, Mean, and Variance

We first work out the m.g.f. which will in turn be used to obtain the mean and variance
expressions.

By definition;
M . g . f . = M X (t )=E (etx )
=E [ e tx +μt−μt ] =E [ e tx−μt⋅e μt ]
= e μt E [ etx −μt ] = e μt E [ e t( x −μ) ]
−( x− μ)2
∞ ∞
1 2
= ∫ e μt e t( x−μ ) f (x ) dx = ∫ e μt e t( x −μ ) e2 σ dx
x =−∞ x=−∞ √ 2 πσ 2
−1 −1
∞ [( x−μ )2−2 σ 2 t (x−μ ) ] ∞ [( x−μ )2−2 tσ 2 (x− μ) ]
μt 1 2σ
2 μt 1 2σ
2
=e ∫ e dx =e ∫ e dx
x=−∞ √ 2 πσ 2 x=−∞ √ 2 πσ 2
Consider from:

( x−μ )2 −2 tσ 2 (x −μ )= ( x −μ )2−2tσ 2 ( x−μ )+σ 4 t 2−σ 4 t 2

2
= [ (x −μ )−σ 2 t ] −σ 4 t 2

Then;

2 4 2
∞
1
−1
{
[ ( x −μ)−σ 2 t ] −σ 4 t2 } ∞
1
−1
{[ (x− μ)−σ 2 t ]2 } σ t
2 σ2 2 2
M X (t )=e μt ∫ e dx =e μt ∫ e2 σ ¿ e2 σ dx
x=−∞ √ 2 πσ 2 x=−∞ √ 2 πσ 2
σ2 t 2 ∞ −1
{[ (x−μ )−σ 2 t ]2} 2 2
σ t ∞ −1
{[( x−μ )−σ 2 t ]2}
1 2 μt+
1 2
=e μt⋅e 2
∫ 2
e2 σ dx =e 2
∫ 2
e2σ dx
x=−∞ √ 2 πσ x=−∞ √2 πσ

Page 6 of 8
 ∞
1
−1
{[( x−μ )−σ 2 t ]2 }
2σ 2
∫ e dx
The integral x=−∞ √2 πσ 2 is 1 since it is the area under a normal
2
distribution with mean μ+σ t 2
and variance = σ .

σ2t2
μt +
2
Hence M X (t ) =e

Next, we know that

Mean = M 'X (0 )

σ 2t2 σ 2 t2
∴ M 'X (t ) =
d μt+ 2
dt
e [ ]=(μ +σ t )e 2
μt+
2

At t = 0; we have

M 'X (0) = ( μ+σ 2⋅0)e 0=μ

⇒ Mean=μ
Next we work out Var (x)

We know that

2 2 2
Var( X ) = σ = E ( X ) − [ E( X )]
'' ' 2
= M X ( 0) − [ M X (0 )]

σ2t 2
μt +
M 'X (t )= 2
( μ+σ t )e 2

Using product rule to differentiate the above expression, we get

Let

Page 7 of 8
 dU
U=( μ+σ 2 t ) ⇒ =σ 2
dt
σ 2 t2 σ 2 t2
μt+ dV μt+
2 2
V =e ⇒ =( μ+σ t )e 2
dt

'' dU dV
∴ M X (t ) = V +U
dt dt
2 2 2 2
σ t σ t
μt+ μt +
2 2 2 2 2
=e ¿ σ +( μ +σ t )⋅( μ+ σ t )e

At t=0

'' 0 2 2 2 0
M X (0 ) = e ¿ σ +( μ+σ ¿ 0 )⋅( μ+σ ¿ 0 )e
2 2
=μ + σ

But

2
Var( X ) = M ''X (0) − [ M 'X ( 0)]
=μ2 +σ 2 −( μ )2
= σ2
Hence
2
Var ( X) = σ which is the variance of the normal variable X.

Page 8 of 8