Module 1
Bernoulli Distribution
Definition
A discrete random variable X is said to follow a Bernoulli distribution if
its probability density is given by,
P(X=x)= px q1-x ,where x=0,1
p- Probability of success, p>0
q- Probability of failure, q=1-p
x- Number of success p is the parameter of distribution
Mean and variance of Bernoulli distribution
1 1 x
Mean=E(x) = 0
xp(x) = ∑ 0 px q1-x
= 0 + p1 q0 = p
Variance = v(x) = E(x2) – (E(x))2
E(x2) = ∑ 1 x2 p(x)
0 =∑ 1
0 x2 px q1-x = 0+ p1 q0 = p
Variance = E(x2) – (E(x))2 = p – p2 = p (1-p) = pq
Properties
1) P is the parameter of the distribution
2) The mean of the distribution is P
3) The variance is pq
4) Variance is less than mean
5) If x1,x2,… … ..xn are ‘n’ independently and identically distributed
Bernoulli variables with parameter p, then their sum follows binomial
distribution with parameters n and p.
Continuous uniform distribution
Definition
A continuous random variable X is said to follow uniform distribution if
its probability density function is
f(x)={1/𝑏
𝑏−𝑎 ; a≤𝑥≤𝑏
0 ; x<a or x>b
Mean
Mean =E(x) = ∫ xx f(x) dx
b 1 1 b x
=∫ a x dx = ∫ a dx
𝑏−𝑎 𝑏−𝑎
𝑥2 b
= 1
a
2
𝑏−𝑎
𝑏2 𝑎2
= 1 −
2 2
𝑏−𝑎
𝑏2 − 𝑎2
= 1 = (𝑏
𝑏 + 𝑎)(𝑏
𝑏 − 𝑎) = a+b
2
𝑏−𝑎 (𝑏 − 𝑎)2 2
Variance
V(x) = E(x2) – (E(x))2
1
E (x2) =∫ x x2 f(x) dx = ∫ b x2
a dx
𝑏−𝑎
1 b 2
= ∫ a x dx
𝑏−𝑎
1 𝑥3 b 1 𝑏 3 − 𝑎3
= [ ] a= ( )
𝑏−𝑎 3 𝑏−𝑎 3
1 (b − a)(b2 + ba + a2)
E(x2) = = (b2 + ba + a2)
𝑏−𝑎 3 3
(b2 + ba + a2) 𝑎2 + 2𝑎
𝑎 𝑏 + 𝑏2
V(x) = – ( )
3 4
𝑏2 + 4𝑎
4𝑏 𝑎2 − 3𝑎
𝑎 𝑏 + 4𝑎 𝑎 2 − 6𝑎 𝑏2
𝑎 𝑏 − 3𝑏
=
12
𝑎2 − 2𝑎
𝑎 𝑏 + 𝑏2
=
12
𝑎 − 𝑏 )2
(𝑎
V(x) =
12
Moment generating function
Mgf = E (etx) =∫ x etx f(x) dx
b etx 1
=∫ a dx
𝑏−𝑎
1 b tx 1 𝑒𝑡𝑥 b
= ∫ a e dx = [ ] a
𝑏−𝑎 𝑏−𝑎 𝑡
1 𝑒𝑡𝑏 − 𝑒𝑡𝑎
= ( )
𝑏−𝑎 𝑡
etb − eta
=
t(b − a)
Properties
𝑎+𝑏
1) Mean of the distribution is
2
𝑎 − 𝑏 )2
(𝑎
2) Variance of the distribution is
12
3) The parameters of the distribution are ‘a’ and ‘b’
𝑒 𝑡𝑏 − 𝑒 𝑡𝑎
4) The mgf of the distribution is
𝑡(𝑏
𝑏 − 𝑎)
Binomial distribution
Definition
A discrete random variable X is said to follow binomial distribution
with parameters ‘n’ and ‘p’ if its probability density function is,
p(x) =n Cx px q n-x , where x=0,1,2,… … .n
where p – probability of success
q – Probability of failure, p + q = 1
x – Number of success
n – Number of trails
Mean
𝑛 𝑛
E(x) = 𝑥=0
𝑥 𝑝(𝑥
𝑥) = 𝑥= 0
𝑥 n Cx px q n-x
n 𝑛!
=∑ x=0 x 𝑥! 𝑛 − 𝑥 ! pxqn-x
( )
n (𝑛
𝑛 − 1) !
= np ∑ x=1 px-1 qn-x
(𝑥 − 1)!(𝑛 − 𝑥)!
= np∑ n Cx-1 px-1qn-x
x=1(n-1)
= np[n-1 C0 p0 qn-1 +(n-1) C1 p1 qn-2 +… … + n-1 Cn-1 pn-1 q0]
= np [qn-1 + (n-1) C1 pqn-2 +… … ..+ pn-1] (p+q=1)
= np [q+p]n-1 = np
Variance
V(x) = E(x2)-(E (x))2
E (x2) = E[x(x-1)+x]
=E[x(x-1)] + E(x)
E[x(x-1)] = ∑ n n Cx px qn-x
x=0 x(x-1)
n
𝑛!
= ∑ x=0 x(x-1) px qn-x
𝑥!(𝑛 − 𝑥)!
n
𝑛!
= ∑ x=0 px qn-x
(𝑥 − 2)!(𝑛 − 𝑥)!
(𝑛 − 2)!
= n(n-1) p2 ∑ n
x=2 px-2 qn-x
(𝑥 − 2)!(𝑛 − 𝑥)!
= n(n-1) p2 [∑ n
x=2 n-2 Cx-2 px-2q n-x]
= n(n-1) p2[n-2C0 p0qn-2 + n-2 C1 pqn-3+… … +n-2 cn-2 pn-2 q0]
= n(n-1) p2 [qn-2 + C1 pqn-3 +… … … + pn-2]
= n (n-1) p2 (q+p)n-2
= n (n-1) p2
E(x2) = E[x(x-)] + E(x)
= n(n-1) p2+ np
V(x) = n(n-1) p2+ np – n2 p2
= n2 p2 – np2 + np – n2 p2 = np – np2
V(x) = np(1-p)
V(x) = npq
Moment generating function
MX(t) = E(etx)
=∑ n tx
x=0 e p(x)
=∑ n tx cxpx qn-x
x=0 e n
=∑ n t x qn-x
x=0 nCx (pe )
= nCo(pet)o qn-0 + n C1(pet)qn-1 +… … ..+ nCn (pet)n qo
= qn+ nC1(pet)qn-1+… … … + (pet)n
= (q+pet)n
Reproductive or additive property of binomial distribution
Let X1and X2 be two independent binomial variables with
parameters (n1,p) and (n2 ,p) respectively .Then X= X1 + X2 follows binomial
distribution with parameters (n1+n2 ,p). This is known as reproductive or
addictive property of binomial distribution.
Proof
X1~b (n1 ,p) and X2 ~ b (n2 ,p)
Mx1 (t) = (q+pet)n1
Mx2 (t) =(q+pet)n2
Mx1+ x2 (t) = Mx1 (t)Mx2 (t)
= (q+pet)n1 (q+pet)n2
= (q +pet)n1+n2
This is the mgf of binomial distribution with parameters (n1+n2, p)
Properties binomial distribution
1) It is a discrete probability distribution.
2) Binomial has one or two model values.
3) Mean of binomial distribution increases as ‘n’ ’ increases with
‘p’ remains constant.
4) Binomial distribution has mean ‘np’ and variance ‘nqp’ .
5) If two independent random variables follow binomial distribution then
their sum also follows binomial distribution.
Problem
1) 4 dice are thrown 162 times. The occurrence of ‘2’ or ‘3’ is
considered as success. In how many throws, do you expect (a) exactly
2 success (b) at least one success.
2 1 2
N= 162, n=4, p=P( success) = P[getting 2 or 3] = = , q=
6 3 3
P[X=x] = nCX px qn-x
a) P[exactly two success] = P[x=2]
1 2 2 4-2 1 4
= 4C2( ) ( ) = 6× ×
3 3 9 9
24
=
81
Therefore number of throws in which exactly two success = P[X=x] ×
24
N= × 162 = 48
81
b) P[at least one success ] = 1-P[ no success]
1 0 2
= 1- P[ x=0] = 1- 4 C0 ( ) ( ) 4-0
3 3
16 65
= 1- =
81 81
Number of throws in which at least one success
65
= × 162 = 130
81
Fitting of binomial distribution
Poisson Distribution
Definition
A discrete random variable X is said to follow Poisson distribution if its
e − mmx
probability density function is, p(x) =
x!
,where x = 0,1,2,… … … .. ‘m’ is the parameter of the distribution.
Mean
Mean = E(x) = ∑ ∞
x=0 x f(x)
∞ 𝑒 − 𝑚 𝑚𝑥
= ∑ x=0 x
𝑥!
∞ 𝑚𝑥 𝑚𝑥 − 1
= 𝑒−𝑚 ∑ x=0 = 𝑒 − 𝑚m ∑ ∞ x=1
(𝑥 − 1)! (𝑥 − 1)!
𝑚0 𝑚1 𝑚2
= m𝑒 − 𝑚 [ + + + … .]
0! 1! 2!
𝑚 𝑚2
= m𝑒 − 𝑚 [ 1+ + +… ..]
1! 2!
= m𝑒 − 𝑚 e m =m
Mean of Poisson distribution is ‘m’ .
Variance
V(x) = E(x) – (E (x))2
E (x2) = E[ x(x-1) +x]
= E [ x (x-1) ] + E(x)
E[x(x-1) ] = ∑ ∞
x=0 x (x+1) f (x)
∞ 𝑒 − 𝑚 𝑚𝑥
=∑ x=0 x (x-1)
𝑥!
𝑚𝑥 − 2
= e-m m2 ∑ ∞ x=2
(𝑥 − 2)!
𝑚 𝑚2
= m2 𝑒 − 𝑚 [1+ + + … ..] = m2 𝑒 − 𝑚 e m = m2
1! 2!
E (x2) = m2 + m
V(x) = m2 + m-m2
=m
s.d = √𝑚
𝑚
Moment generating function (mgf)
Mgf = mx (t) = E [etx]
=∑ ∞ tx
x=0 e f (x)
∞ tx 𝑒 − 𝑚 𝑚𝑥
=∑ x=0 e
𝑥!
(𝑚𝑒 𝑡)𝑥 𝑚 𝑒𝑡 (𝑚 𝑒𝑡 )2
= 𝑒 − 𝑚∑ ∞ x=0
𝑥!
= -m
e [1+
1!
+
2!
+ … … ..]
Mgf = 𝑒𝑚𝑒𝑡 − 1
Poisson distribution as a limiting case of binomial distribution
Binomial distribution tends to Poisson distribution under the following
conditions,
1) Number of trials is very large (n→ ∞)
2) Probability of success for each trials is very small (𝑝→ 0)
3) ‘np’ is finite (ie, np = m )
Therefore distribution can be stated as follows; Let the probability for
success in a single trial of a random experiment is very small (p→0) a
nd the experiment be repeated larger number of times (n→∞). Then
the probability for ‘x’ successes out of the ‘n’ trials is given by
𝑒 − 𝑚 𝑚𝑥
P(x) = .This law is called Poisson law.
𝑥!
Additive or reproductive property of thee Poisson distribution
Let X1 and X2 be the two independent random variables following
Poisson distribution with parameters m1and m2 respectively .Then
their sum is also follows Poisson distribution with parameter m1 + m2.
Proof
Let X = X1 + X2
MX(t) = MX1+X2 (t) = Mx1 (t) Mx2 (t)
= 𝑒𝑚1𝑒
𝑒𝑡 − 1 𝑒 𝑚2𝑒
𝑒𝑡 − 1
= 𝑒(𝑚
𝑚1 + 𝑚2)𝑒
𝑒𝑡 − 1
This is the mgf of Poisson distribution with parameter ‘m1+m2’
Properties of poisson distribution
1) It is a discrete probability distribution
2) Poisson distribution has a single parameter ,m
3) Mean and variance of Poisson distribution are equal to ‘m’
4) If ‘X’ follows a Poisson distribution, then ‘X’ can take values
0,1,2,… … to infinity .
Problems
1) Out of 500 items selected for inspection 0.2% are found to be
defective. Find how many lots will contain exactly no defective if
there are 1000 lots.
n=500, p = 0.2% = 0.002
N = 1000, np = 500 × 0.002 = 1
Therefore , m = 1
𝑒 − 𝑚 𝑚𝑥 𝑒 − 1 1𝑥
P(x) = =
𝑥! 𝑥!
𝑒 110
P [x=0] = = 0.3679
0!
Therefore number of lots containing no defective =P [ x=0] N
= 1000 × 0.3679
= 367.9 = 368
Fitting of Poisson distribution
Normal distribution
Definition
A continuous random variable X is said to follow normal distribution
− (x − μ)2
1
with parameters µ and 𝜎 if its pdf is given by, f(x) = e 2σ2
σ √2π
; -∞ ≤ 𝑥 ≤ ∞
Where 𝜎 is the mean and 𝜎 is the s.d. of the distribution
Normal curve
The graph is bell shaped and is symmetric.
Properties of normal distribution (normal curve)
1) The normal curve is bell shaped.
2) The normal curve is continuous curve.
3) Normal curve is symmetric about the mean.
4) Mean, median and mode are equal for normal distribution.
5) The height of normal curve is at its maximum at the mean.
6) There is only one maximum point which occurs at the mean.
7) The ordinate at mean divides the whole area into two equal parts.
8) The curve is asymptotic to the base on either side i.e, the curve
approaches nearer and nearer to the base, but it never touches.
9) The coefficient of skewness is zero (𝛽1=0).
10) The normal curve is unimodel. i.e it has only one mode.
11) Q1 and Q3 are equidistant from median.
2
12) Mean deviation for normal distribution is 4 𝜎 and Q.D = 𝜎
3
5
13) No portion of the curve lies below X-axis
14) All odd moments of normal distribution are zero and even
moments are interrelated,
µ2n = (2n-1) 𝜎2 µ2n-2
15) Normal distribution is mesokurtic. ie β2= 3
16) If X and Y are two independent normal variates then their sum is
also a normal variate.
Standard normal distribution
If X is a random variable following normal distribution with mean µ
and standard deviation 𝜎 then the variable Z = 𝑥− µ is known as standard
𝜎
normal variate. This Z follows normal distribution with mean zero and s.d =
1. The distribution of Z is known as standard normal distribution. The
1
probability function of Z is f(z) = 𝑒𝑧 2 /2 ; -∞ < 𝑧 < ∞
√2𝜋
𝜋
Mean of normal distribution
Mean = E(x) = ∫ xx f(x)dx
= ∫ ∞ (x- µ+ µ) f(x) dx
-∞
− (x − μ)2
∞ ( 1
= ∫ -∞ x- µ+ µ) e 2σ2
𝜎√2𝜋
𝜋
− (x − μ)2
∞ (x- 1 ∞
= ∫ -∞ µ) e 2σ2 dx + ∫ -∞ μ
𝜎√2𝜋
𝜋
− (x − μ)2
1
𝜎√2𝜋
𝜋
e 2σ2 dx
− (x − μ)2
µ ∞
= 0+ ∫ -∞ e 2σ2 dx
𝜎√2𝜋𝜋
− (x − μ)2
µ (x − μ)2
Mean = 2∫ μ
∞ e 2σ2 dx let y=
𝜎√2𝜋
𝜋 2σ2
2µ ∞ 𝜎2𝑑𝑦 = 2(x − μ)dx
= ∫ 0 e-y dy
𝜎√2𝜋
𝜋 𝜎2 𝑦
2𝜎 2σ2
𝜎2
2µ𝜎 1 𝑒 −𝑦 σ2 dy
= ∫ ∞
0 dy dx=
𝜎√2𝜋𝜋 √2𝜎
𝜎 √𝑦
𝑦 x−μ
µ ∞ e−y
= ∫ 0 y1/2 dy 2σ2y = (x- μ)2
√𝜋
𝜋
= µ ∫ ∞ dy x- μ = 2𝜎
𝜎2 𝑦
0 𝑒−𝑦 𝑦 − 1/2
√𝜋
𝜋
µ ∞ 𝑒 − 𝑦 y1/2-1 dy 𝜎2𝑑𝑦
= ∫ 0 dx =
√𝜋
𝜋 2𝜎
𝜎2𝑦
= µ 1/2 x = μ, y=0
√𝜋
𝜋
= µ 𝜋 =µ x= ∞, y=∞
√𝜋
𝜋
Therefore mean = µ
∞
Note: ┌p = 𝑒 − 𝑥 𝑥𝑝 − 1 𝑑𝑥
0
┌p = (p-1) ┌(p-1), ┌p = (p-1)!
1/2 = 𝜋
Variance
µr = E [X - mean]r = E [X − µ]r
r= 2K+1 for odd moments
Therefore µ2k+1 = 0
2
M2K= E[X-µ]2K let y= (x − μ)
2σ2
=∫ ∞ (X-
-∞ µ)2K f(x)dx dy = 2(x − μ)dx
2σ2
− (x − μ)2
M2K = ∫ ∞ (x
-∞ - µ)2k 1
e 2σ2 dx dx= σ2 dy
𝜎 2𝜋
𝜋 x−μ
− (x − μ)2
= 2
∫ ∞ (x- µ)2k e dx 2σ2y = (x- μ)2
μ 2σ2
𝜎 2𝜋
𝜋
= 2
∫ ∞ ( 2σ2 y ) 2K e-y 𝜎2𝑑𝑦
x- μ = 2𝜎
𝜎2 𝑦
0
𝜎 2𝜋
𝜋 𝜎2𝑦
2𝜎
= 2 𝜎2
∫ ∞ ( (2𝜎
𝜎 2 𝑦 )1/2 )2k 𝑒 − 𝑦𝑑𝑦 dx =
0
𝜎 2𝜋
𝜋 √2𝜎
𝜎 √𝑦𝑦
𝜎2 𝑑𝑦
𝜎2𝑦
2𝜎
1 ∞
=
√𝜋
𝜋
∫ 0 𝜎2 )k yk e-y y-1/2 dy
(2𝜎 x = μ, y=0
2𝑘 𝜎2𝑘
𝑘
∞ e-y yk-1/2 dy
= ∫ 0 x= ∞, y=∞
√𝜋𝜋
2𝑘 𝜎2𝑘
𝑘
∞ e-y yk+1/2 - 1
= ∫ 0 dy
√𝜋
𝜋
2𝑘 𝜎2𝑘
𝑘 1
M2k =
√𝜋
𝜋
┌(k+ 2
)
Put k = 1
𝜎2
2𝜎
M2 =
√𝜋
𝜋
┌(1+ 1
)
2
1
= 2𝜎2 ┌3/2 = 2𝜎2
2
┌1/2
𝜋 𝜋
= 𝜎2 √ 𝜋 = 𝜎2
𝜋
Therefore variance = 𝜎2
Mgf
Mx (t) = E [etx]
∞ x− μ
=∫ -∞ etx f(x) dx z=
σ
, x= μ + zσ
− (x − μ)2
∞ tx 1 1
=∫ -∞ e e 2σ2 dx dz = dx , dx = σ dz
𝜎 2𝜋
𝜋 σ
− (x − μ)2
1 ∞
= ∫ -∞ etx e 2σ2 dx x = -∞, z = -∞ and x= ∞, z=
𝜎 2𝜋
𝜋
1 +∞
= −∞
e t(μ + zσ) e − z2/2 σdz
𝜎 2𝜋
𝜋
𝜎 z2
= eμt ∫ ∞ etσz − dz
-∞ 2
𝜎 2𝜋
𝜋
𝑒μ𝑡𝑡 2tσz − z2
= ∫ ∞ dz
-∞ e 2
2𝜋
𝜋
eμt ∞
= ∫ 2
-∞ e − 1/2[z − 2tσz] dz
2π
eμt
= ∫ ∞ 2 2 2 2 2 dz
-∞e − 1/2[z − 2tσz + t σ − t σ ]
2π
eμt t2σ2
= ∫ ∞ 2 2 2 dz
-∞e − 1/2[z − 2tσz + t σ ] e 2
2π
𝑡2σ2
𝑒µt 𝑒 2
= ∫ ∞ dz
-∞ e − 1/2(z − tσ)
2
2𝜋
𝜋
t2σ2
µt +
𝑒 2
∞
= ∫ -∞ 𝑒 − 1/2(𝑧 − 𝑡𝜎)
2 dz
2𝜋
𝜋
𝑡2σ2 1
µt + ∞ 𝑒 − 1/2(𝑧 − 𝑡 𝜎)2
=𝑒 2 ( ∫ -∞ dz =1. Pdf of standard normal
2𝜋
𝜋
distribution).
Problems
1) The weekly wages of 1000 workmen are normally distributed with mean
Rs. 70 and s.d of Rs 5. Estimate the number of workers whose weekly wages
will be
a) Between 70and 72
b) More than 75
c) Less than 63
Here, µ=70,𝜎 = 5
70 − 70 𝑥−µ 72 − 70
a) P[70<x<72] = P [ < < ]
50 𝜎 50
= P [0<z<0.4] = 0.1554
Number of workers = 1000 × 0.1554 = 155.4 ≈ 155
𝑥−µ 75 − µ
b) P [x>75] = P [ > ]
𝜎 𝜎
75 − 70
= P [z> ] = P [z>1]
5
= P [1<z<∞]
= P [0<z<∞] − P [0<z<1]
= 0.5 – 0.3413= 0.1587
Therefore number of workers = 1000 × 0.1587 = 158.7 ≈ 159
𝑥−µ 63 − µ 63 − 70
c) P[x<63] = P[ < ] = P[z< ]
𝜎 𝜎 5
= P[z<-1.4]
=P[-∞ < z<0] - P[0<z<1.4]
=0.5-0.0808
= 0.4192
Number of workers = 1000 × 0.4192 = 419.2 ≈ 419
2) In a normal distribution 17% of the items are below 30and 17% of the
items are above 60. Find the mean and s.d
17% of item are below 30 means,
P[x<30] = 17% = 0.17
𝑥−µ 30 − µ
P[
𝜎
< 𝜎
] = 0.17
P [z<a] = 0.17
P [0<z<a] = 0.5-0.17=0.33
a = -0.95
30 − µ
= -0.95
𝜎
30 − µ = -0.95 𝜎 → (1)
17% of area above 60
P[x>60] = 0.17
𝑥−µ 60 − µ
P[ > = 0.17
𝜎 𝜎
P [z>b] = 0.17
P [0<z<b] = 0.5-0.17=0.33
∴ b = 0.95
60 − µ
= 0.95
σ
60 − µ = 0.95σ → (2)
(1)+ (2) =>90-2 µ = 0
90=2 µ
µ = 45
(2)=> 60-45 = 0.95 𝜎
15 = 0.95 𝜎
𝜎= 15 = 15.78
0.95
∴ µ = 45, 𝜎 = 15.78