CH 4 For Display

BITS Pilani
Pilani Campus
Chapter 4
Continuous Distribution
Continuous Random Variables
Definition: A random variable is continuous
if it can assume any real numbers and some
interval (or intervals) of real numbers and
the probability that it assume any specific
value is 0 (zero).
BITS Pilani, Pilani Campus

CONTINUOUS DENSITY (Probability
density function )
Definition: Let X be a continuous random
variable. A function f(x) is called continuous
density (probability density function i.e. pdf )
iff 1. f(x)  0

2.  f ( x ) dx  1

b
3.P[ a  X  b]  
a
f ( x ) dx
for any a  b.
Note :
 f ( x)dx  k  1,0  k  


Probability for Interval
1.PX = a] =0.
2. PX = b] =0.
3. P[a  X b] = P[ a  X < b]
= P[a < X  b] = P[a < X < b]
b
4. P[a  X b] =

a
f ( x)dx;
where a and b are real numbers. Area under the
curve of f between x = a to x = b. TOTAL AREA IS
ONE. BITS Pilani, Pilani Campus
CUMULATIVE DISTRIBUTION FUNCTION
Let X be the continuous r.v. with density f(x).
The cumulative distribution function (cdf) for X,
denoted by F(X) , is defined by
F(X) = P ( X  x ) , all x
x
= 

f (t ) dt

PROBABILITY by using cdf F(x)
P(a  X  b) = F(b) – F(a).
FIND f(x) from F: f(x) = dF(x)/dx = F(x)
for all x.
Note:
1. If X is a Continuous random variable, then
F(x) is also a continuous function for all x.
2. The F(x) is nondecreasing function.
3. lim F ( x )
x  
x 
 lim
x   

f (t ) dt  

f (t ) dt  0
4. lim F ( x)
x  
x 
 lim
x   

f (t )dt  

f (t )dt 1
Note : F is cdf of continuous rv if and only if 1., 2, 3.

and 4. hold.
EXERCISE 4.1, 9 PAGE 140
Consider the density of X
f (x ) = x / 6 ; 2  x  4 ,
= 0 ; e.w.
(a) Find the cumulative distribution function.
(b) Use cdf F, find P[2.5X 3], P[1 X3.5].

Example
The Cumulative distribution function (c.d.f.)
of a continuous random variable X is
 0, x0
 1
a  bx 2 , 0 x
 2
F ( x)  
1  3 3  x 2 , 1
 x3
 25 2
1 , x3

Determine a and b. Hence find the
probability density function (p.d.f.).
EXERCISE 4.1.14, PAGE 141
In parts (a) and (b) proposed cumulative
distribution functions are given. In each case,
find the density that would be associated with
each, and decide whether it really does define
a valid continuous density. If it does not,
explain what property fails.
(a) Consider the function F defined by
F(x) = 0 ; x  -1,
= x + 1 ; -1  x  0,
= 1 ; x > 0.
0, x  1
dF 
f ( x)   1,  1  x  0
dx 
0, x0
Clearly f(x) 0 for all x. Also F is cdf as



f ( x)dx  1

…. continue
4.1.14 (b)
Consider the function defined by
F(x) = 0 ; x0 ,
= x2 ; 0 < x  0.5 ,
= x/2 ; 0.5 < x  1 ,
= 1 ; x> 1 .

dF
Check for f ( x)  ,  f ( x)dx  1.
dx 
(5) (Continuous uniform distribution) A
random variable X is said to be
uniformly distributed over an interval
(a, c) if its density is given by
1
f ( x)  , axc
ca
(a) Show that this is a density for a
continuous random variable.
1
Sol: Since  0 , for c  a
ca BITS Pilani, Pilani Campus
Secondly,
c
1
a c  a dx  1
(b) Sketch the graph of the uniform density.
f(x)
1/(c-a)
x
X=a X=c
( ii ) Shade the area in the graph of part (b)
that represents P[X  (a + c)/2].
f(x)
(a+c)/2 X= c x
X= a
(c ) Find the probability pictured in part: ii
a c
 a  c 2
1
P X 
 2   
a
ca
dt  0.5
(e) Let (l, m) and (d, f) be subintervals of (a, c)

of equal length. What is the relationship
between P[l  X  m] and P[d  X  f]
Sol: Probability is same on equal
length of interval.
(10) Find the general expression for the
cumulative uniform distribution for
a random variable X over (a, c)
F ( x)  0, xa
x
1
F ( x)  P[ X  x]   ds
a
( c  a )
xa
 , a xc
ca
F ( x)  1, x  c
EXPECTATION & DISTRIBUTION
PARAMETERS
 xf ( x ) dx


2. Let X be the c.r.v. with density
f(x). Let H(x) be a random
variable. The expected value of
H(x) is defined as:
E H  X    H ( x ) f ( x ) dx
provided  


| H ( x ) | f ( x ) dx
is finite.
…continue
3. Var X = E[X2] – (E[X])2 = 2 , where

E[ X ]   x f ( x)dx
2 2

Variance is shape parameter in the
sense that a random variable with small
variance will have a compact density;
one with a large variance will have a
density that is rather spread out or flat.
…continue
4. m.g.f. = Moment Generating Function



= E[ etX] = mX (t) = tx
e f ( x ) dx


Section 4.2, page 141, 17
Let X denote the length in minutes of a long

distance telephone conversation. The
density for X is given by
x
1 
f ( x)  e 10
, x0
10
(a)Find the moment generation function
mX(t).

MGF of uniform distribution
random variable on (a, c) :
c
1
E[e ]   e
tX tx
dx
a
c  a
1 e  e
tc ta

  , t  0
t  ca 
Use definition to find mean and
variance can be found.
ac
E( X ) 
2
(c  a ) 2
Var ( X ) 
12
Section 4.1, page no 139, 6
If a pair of coils were placed around a

homing pigeon and magnetic field was
applied that reverses the earth’s field, it is
thought that the bird would become
disoriented. Under these circumstances it is
just as likely to fly in one direction as in any
other.

q
Pigeon
Home (0)
q is uniformly distributed over the interval [0, 2p].
a) Find the density for q.

(b) Sketch the graph of the density.
(c) Shade the area corresponding to the
probability that a bird will orient within
p/4 radians of home, and find this area
using plane geometry.
(d) Find the probability that a bird will orient
within p/4 radians of home by integrating
the density over the appropriate
region(s).
(e) If 10 birds are released independently
and at least 7 orient within p/4 radians of
home, would you suspect that perhaps
the coil are not disorienting the birds to
the extent expected? Explain, based on
the probability of this occurring.
(f) Find mean, variance for q.

Example : A random variable X with density
1 a
f (x)  ,
p a  ( x  b)
2 2
   x  ,    b  , a  0
is said to have a Cauchy distribution with
parameters a and b. This distribution is
interesting in that it provides an example of a
continuous random variable whose mean does
not exist. Let a = 1, b = 0 to obtain a special
Cauchy distribution with density
1 1
f ( x)  ,   x  
p 1 x 2

Show that  | x | f ( x) dx does not exist.

Sol:
 
1 1
| x | f ( x ) dx   p 1  x 2
| x | dx


1 x
0
1 x
  dx   dx
p 1  x 2
p 0 1 x 2
Multiply and divide by 2, we get


 1
0
2  1 2 
  ln | 1  x |   ln | 1  x |
 2p    2p 0
which does not exist, as
ln( x)   as x  .
GAMMA DISTRIBUTION
A random variable X with density function
1 a 1  x / b
f ( x)  x e , x  0,
(a ) b a
=0, for x  0,
is said to have a Gamma Distribution with
parameters a and b,for a > 0,b > 0.

Gamma Function
Definition : The function


 defined by
(a)   e z dz; z  0, a  0,
 z a 1
0
is called the Gamma function.
Properties :
1. (1) = 1
2.  (a) = (a -1)  (a -1 ); where a>1 is real
number.
3. (n) = (n-1)! (Factorial of n-1), n = 1,2,3….
4.  (1/2)=p
1. (1) = 1
Proof: By definition of Gamma function, we
have 
(1)   e z dz; z  0, a  1,
z 0
0

  e dz  1
z

2. (a) = (a -1)  (a -1 ); where a is the +ve
real number.
Proof: 
(a)   e z dz; z  0, a  0,
 z a 1
by integrating by parts, we have


 z
 (a )   e z 
a 1 
0  (a  1)  e  z z ( a 1)1dz
0

a 1
z a 1 z
lim ( e z )  lim z
z  z  e
a2
 (a  1) z
 lim z
z  e
a 3
 (a  1)( a  2) z
 lim z
z  e
 1 
 (a  1)! lim  z   0

z  e

Thus, 
(a)   e z
z 
a 1 
0  (a  1)  e z
z ( a 1) 1
dz
0
 (a  1)(a  1)
3. (a)  (a  1)!
Since,
(a)  (a  1)(a  1)  ...
 (a  1)(a  2)...(1)  (a  1)!

1
1

    z e  z dz  p
2
2 0

1
   2 e dt
t 2
2 0
Convert to polar coordinates and see it is π.

GAMMA DISTRIBUTION
 1
 x e , x  0,a  0, b  0
a 1  x / b
f ( x)   (a ) b a
0,
 other wise

parameters a and b,for x > 0,a > 0,b > 0.

To check the necessary and
sufficient condition of pdf:
f ( x )  0 for all x  0
 
1
Further,  f ( x) dx  a x
a 1
e x /b
dx

 (a ) b 0
x
Let  t  dx  b dt , and x  b t
b

1
a 
a 1 a 1 t
 b t e b dt
 (a ) b 0
 a 
b
 a 
a 1  z
f ( x ) dx  z e  1
 (a ) b 0
Hence f ( x) is a pdf

MEAN, VARIANCE AND MGF OF GAMMA FUNCTION
Let X be a Gamma r.v. with parameters

aand b.
Then 
1 x
0 (a )b a
a 1
x exp(  ) dx  1
b
1. m.g.f. = mX(t)=(1- bt)-a , t < 1/ b
2. E[X] = Mean = X = a b
3. Var(X) = 2 = a b2
Proof: By def , mX (t )  E[etx ]

1
 e tx
x a 1  x / b
e dx
0
(a ) b a
 (
1
 t)x
1
x
a 1 b
 e dx
(a ) b a
0
x zb
let z  (1  bt )  x 
b (1  b t )
bdz
and dx 
(1  bt )

1 bz a 1  z bdz
a 
 ( ) e
 (a ) b 0 1  b t (1  bt )
a 
1 b
a 
a 1  z
 a
z e dz
(a) b (1  bt ) 0

a
1 b
m X (t )  a a
 ( a )
(a) b (1  bt )
a 1
m X (t )  (1  bt ) , t
b

d
E[ X ]  mX (t )
dt t 0
 a 1
  a(1  bt ) (b)  ab
t 0

2
d
E[ X ]  2 ( m X (t ))
2
dt t 0
 a(a  1)b 2
Var ( X )  E[ X ]  E[ X ] 
2 2
 a(a  1)b  (ab)  ab

2 2 2

• a and b both play a role in determining
the mean and the variance of the
random variable.
• Graphs of Gamma Densities are not
symmetric and are located entirely to
the right of the vertical axis.

0.25
Gamma(1, 4)
f(x) 0.2
Gamma(2, 3)
0.15 Gamma(20, 0.5)
0.1
0.05
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
x
For a > 1, the maximum value of the
density occurs at the point x = (a – 1)b.

0.25
Gamma(1, 4)
f(x) 0.2
Gamma(2, 3)
0.15 Gamma(20, 0.5)
0.1
0.05
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
x
Exercise 4.3 (Page No: 143)
Let X be a gamma random variable with
a = 3 and b = 4
(i) What is the expression for the density
for X x x
1 a 1 b 1 2 4
f ( x)  a
x e  x e , x  0
 (a ) b 2!4 3
x
1 2
 x e , x0
4
128
(ii) What is the moment generating
function for X
a 3 1
m X (t )  (1  bt )  (1  4t ) , t 
4
(iii) Find mean, variance and standard
deviation
  ab  12
  ab  48
2 2
  48  6.9282
Exponential Distribution
In Gamma Distribution, put a = 1, we
get
1 x b
 e , x  0, b  0
f ( x)   b
0 ,
 elsewhere

0.3
0.25
0.2
f(x)
0.15
0.1
b3
0.05
0
1 3 5 7 9 11 13 15 17 19
x
Cont…
For the above function

Mean =  = b, var (X) = b2,
mX(t)=(1- bt)-1 , t < 1/b

The c.d.f. of exponential distribution
with Parameter b is given by
x
s

x 
s b
1 1 e
F ( x)   e b
ds 
b b  1
0
  
 b 0
s x x
 
b b
 e  1 e x0
0
The c.d.f. of exponential
distribution
0, x0

F ( x)   
x

1  e
b
x0

Poisson Process and Exponential
dist :
Theorem 4.3.3 For a Poisson
process with parameter , Let W be
the time of the occurrence of the
first event. W has an exponential
distribution with parameter b = 1/.

Proof:
The distribution function F for W is given by

F( w )  P[ W  w ] 1  P[ W  w ]
Here, we have that the first occurrence of
the event will take place after time w only
if number of occurrences in the time
interval [0,w] is zero.
Let X be the number of occurrences of
the event in this time interval [0,w].
X is a Poisson random variable with
parameter w. Thus,
P[W  w]  P[ X  0]
 w
e (w)0
 w
 e
0!
 F ( w) 1  P[W  w]  1  e  w

Since, in the continuous case, the
derivative of the cumulative distribution
function is the density
F (w)  f ( w)  e  w
This is exactly density for an exponential

1
random variable with β 
λ
Section 4.3, page no 144,34
A particular nuclear plant releases
a detectable amount of radioactive
gases twice a month on average.
Find the probability that at least 3
months will elapse before the
release of first detectible emission.
What is the average time one must
wait to observe the first emission?
Solution :
Let X be time elapsed before the
release of the first detectable emission
a then X is an exponential distribution
with b = ½, as the radioactive gas is
released in Poisson process with =2.
2 x
Therefore, f ( x)  2 e ,x 0
6 6
P[ X  3] 1  F (3) 1  (1  e )  e ,
Example
A computer centre maintains a telephone
consulting service to trouble shoot for its
users. The service is available for 9:00 to
5:00 each working day. Past experience
shows number of calls received per day is
a Poisson distribution with parameter
50. For a given day find the probability that
first call will be received
(i) before 10:00 a.m. (ii) after 3:00 p.m.
GAMMA RANDOM VARIABLE

 1
 x e , x  0,a  0, b  0
a 1  x / b
f ( x)   (a ) b a
0,
 other wise

parameters a and b,for x > 0, a > 0, b > 0.

Chi-square distribution
If a random variable X has a gamma

distribution with parameters b = 2 and
a = /2,  is a positive integer, then X is
said to have a chi-square (2) distribution
with2  degrees of Freedom and denoted
by   .

b = 2 and a = /2,

1 1
x/ 2
f ( x)  2
x e ,x 0
   /2
 2
2
E[   ] = , Var[   ] = 2 
2 2

Chi square distribution for  1
 1
 x e , x  0,
(1/ 2 ) 1  x / 2
f ( x)   (1 / 2) 2
f(x)
1/ 2
0,
 other wise

Chi square distribution

1 1
x / 2
f ( x)  x 2
e ,x 0
 /2
( ) 2
2
for   2
1 x / 2
f ( x)  e , x0
2
Chi square distribution for  2
1 x / 2
f(x) f ( x)  e , x  0
2

for  = 10
0.1
f(x)
0.05
-1.94E-16
0 5 10 15 20 25
x
We do not have explicit formula for CDF
F of X2 . Instead values are tabulated on
page no. 695-696 as below (F occurs in
margin here, and related value of r.v.
inside the table):

P[X2 < t]
 F 0.100 0.250 0.500
5 1.61 2.67 4.35

6 2.20 3.45 5.35
7 2.83 4.25 6.35
If F is CDF for Chi square random variable

Having 5 degrees of freedom
F(1.61) = 0.1, F(4.35) = 0.5
Another notation
For 0 < α < 1, we denote by  , for a
2
a
chi-square r.v. with  degrees of
freedom, a unique number such that
≥
2
2
P[X  a] =α
Probability to the right of 
2
a is α.
for exampleP[X210 ≥  2
0.05 ] = 0.05
Probability to the right of  is α.
2
a
0.1
f(x)
0.05
0.05
0
0 5 10 15 20 25
P[X210 ≥
2
0.05 ] = 0.05 x
= 1 - P[X210 ≥
2
0.95 ] = 0.95
P[X2  ≤ t]
 F 0.90 0.95 0.975
10 16.0 18.3 20.5

11 17.3 19.7 21.9
12 18.5 21.0 23.3

Probability to the right of  is r.
2
r
0.1
f(x)
0.05
0.05
-1.94E-16
0 5 10 15 20 25
For 10 degrees of freedom,  2
.05  18.3 x
P[X210 ≥
2
0.05 ] = 0.05
38. Consider a chi – squared random
variable with 15 degrees of freedom.
2
(i) What is the mean of X 15 ?
Sol: Chi square distribution is Gamma
with a = /2 & b = 2, Mean = ab =  =15
and 2 = ab2 = 30
(ii) What is the expression for the density for
2
X ?
15
Sol:  is random var iable with

2

b  2, a   / 2, hence
1 (15 / 2 ) 1  x / 2
f ( x)  x e ,x 0
(15 / 2) 215 / 2
0 otherwise

(iii) What is the expression for the
moment Generating function for
2
X 15
Sol:
a
mX (t )  (1  bt )  (1  2t ) , t  1 / 2
15 / 2

(iv) Find P[ X  22.3]
2
15
P[X2 < t]
 F 0.005 0.010 .025 0.900
15 4.60 5.23 6.26 22.3

16 5.14 5.81 6.91 23.5
17 5.70 6.41 7.56 24.8
2
P[ X 15  22.3]  1  P[ X 15
2
 22.3]
 1  0.900  0.10
(v) Find P[6.26  X  22.3]
2
15
P[X2 < t]
 F 0.005 0.010 .025 0.900
15 4.60 5.23 6.26 22.3

16 5.14 5.81 6.91 23.5
17 5.70 6.41 7.56 24.8
P[6.26  X 15  22.3]
2
 F (22.3)  F (6.26)  0.900  0.025  0.875

(vi) Find 20.05, & 20.01 for
15 degree of freedom
Sol. P[X215 > 2r] = r.
P[X2 < t]
 F 0.005 0.010 .025 0.950
15 4.60 5.23 6.26 25.0
P[X215 > 20.05] = 0.05.

P[X215 ≤ 20.05] = 0.95.
 2
0.05
 25.0 for15 degree of freedom
P[X2 < t]
 F 0.005 0.010 .025 0.990
15 4.60 5.23 6.26 30.6
P[X 2 2
> r] = r.
15
P[X215 > 20.01] = 0.01.
P[X215 ≤ 20.01] = 0.99.
 0.01  30.6
2
4.4 The Normal Distribution
A random variable X with density f(x)
is said to have normal distribution
with parameters  and  > 0, where
f(x) is given by:

 x   2
1 2 2
f ( x)  e , x,  (, );   0.
2p 
(i) f ( x)  0 x  (-, )

(ii)  f(x)dx  1 2
-  1  x  
1   
e
2  
To prove dx  1
2p - 2
 x  
let    z  dx  dz
  
2
 1  x    1
1    1  z2
e 
2  
 dx  e 2
dz
2p 2
- 2p -
 1  1
1  z2 1  z2
 e 2
dz  2  e 2
dz
2p - 2p 0
 1  1
 z2  z2
 e dz   e dz  I
0
2
0
2
   
 1  1
 x2  y2
I  I  e 2
dx   e 2
dy 
0  0 
  

1
 ( x2  y2 )
  e 2
dxdy 
0 0  BITS Pilani, Pilani Campus
  

1
 ( x2  y2 )
 e dxdy 
2
0 0 
 lim  p /2 R 1 2

 (r )
q
   e rdrd
2

R 
0 0 
 p /2 R /2

2
 lim    e w
dw dq 
R 
0 0 
p

2 BITS Pilani, Pilani Campus
p
i.e. I 
2
 1  1
1  z2 1  z2
 e 2
dz  2  e 2
dz
2p - 2p 0
2 p
  1
p 2
Mean and Standard deviation
for Normal distribution
Theorem: Let X be a normal
random variable with parameters 
and . Then  is the mean of X and
 is its standard deviation.
Proof deferred.

The density Curves

Cumulative Distribution
Function

Standard Normal Distribution
A random variable Z is called standard normal

if it is normal and with parameters 0,1.

x 
 s  2
1

2 2
F ( x)  P( X  x)  e ds,
2p  
s 
Let z  s    z  ds  dz
 x
z2
1  
 x-  x- 
F ( x)   dz  P Z    FZ  
2
e
2p       
Thus the cdf of (X-μ)/σ is the cdf of standard
normal distribution.
Rajiv
Moment Generating Function
 z2
1 
mZ (t )  E[e ]  
Zt zt 2
e e dz
2p 
 z2
1 zt 
 
2
e dz
2p 

2
z  z  2tz  t
2 2
 2
t t 2
 ( z  t )2 
zt         
2  2  2 2  2 
 ( z t ) 2 
t2
   
1 
mZ (t )  e 
2 p 
2
e  2 
dz
Let z - t = w  dz = dw
 w2 
t2
    t2
1 
mZ (t )  e 2
2p e

 2 
dw  e 2

Let Z be normally distributed with
parameters  = 0 and  = 1, then
the moment generating function for
Z is 2
t
m (t )  e
Z
2

Let X be normally distributed with
parameters  and , then the moment
generating function for X is given by
t(Z   )
m X (t )  E[e ]  E[e
Xt
]
t 2 2
t ( t ) Z t
 e E[e ]e e 2
t 2 2
t 
e 2
Mean and Variance for Normal r.v.
d
Mean  E[ X ]  m X (t )  
dt t 0
2
d
and E[X ]  2 m X (t )    
2 2 2
dt t 0
Var ( X )  E[X ] - E[X]      

2 2 2 2 2
 2
(i) If X is N(, )
 c   a  
P(a  X  c)  FZ    FZ  
     
(ii) FZ ( z)  1  FZ ( z)
(iii ) FZ ( z )  0 if z  -3.5
FZ ( z )  1 if z  3.5
Symmetry
F(2μ-a)=1-F(a).
a-μ
2μ-a μ a

Example
Using table, find the values of
(v) P[-1.305 ≤ Z ≤ 1.43]
(vi) z .10 (vii) z .90
(viii) The point z such that
P(-z  Z  z) = 0.9

(v) P[-1.305 ≤ Z ≤ 1.43]
z 0.00 0.01 0.02 0.03 0.04
1.3 .9032 .9049 .9066 .9082 .9099

1.4 .9192 .9207 .9222 .9236 .9251
P[-1.305 ≤ Z ≤ 1.43] = F(1.43) - F(-1.305)

= 0.9236 – (1 – F(1.305))
= 0.9236 – 1+(0.9032 + 0.9049)/2)
= 0.82765
(vi) z .10 (vii) z .90
(vi) P(Z ≥ zr ) = r
 P(Z ≥ z.10 ) = 0.10
 P(Z < z.10 ) = 0.90
 z.10 = 1.28
z 0.05 0.06 0.07 0.08 0.09

1.2 .8997 .9015
1.4

(vii) z .90 = ?
(viii) The point z such that
P(-z  Z  z) = 0.9
P(-z  Z  z) = F(z) - F(-z)
=2F(z) – 1 = 0.9
i.e. F(z) = 0.95
 z = 1.645
z 0.04 0.05
1.6 0.9495 0.9505

1.4
Example
X is normally distributed and the mean of X is
12 and standard deviation is 4.
(i) Find P(X  10), P(X  10), P(0  X  12)
(ii) Find r, where P(X > r) = 0.24
(iii) Find a and b, where P(a < X< b) = 0.50
and P(X > b) = 0.25

(i) Find P(X  10), P(X  10), P(0  X  12)
 X -  10  12 
P(X  10)  P  
  4 
 P(Z  -0.5)  1 - P(Z   0.5)
 1 - F(-0.5)  1  0.3085
P(X  10) = 1 - P(X > 10) = 0.3085

 0 - 12 X -  12  12 
P(0  X  12)  P   
 4  4 
 F(0) - F(-3)  0.5  0.0013
 0.4987

(ii) Find r, where P(X > r) = 0.24
 X -  r  12 
P(X  r )  P    0.24
  4 
 r  12   r  12 
 P Z    1 - P z    0.24
 4   4 
 r  12 
 F   1  0.24  0.76
 4 
r  12
  0.71 (tabular v alue)
4 BITS Pilani, Pilani Campus
r = 0.71  4 + 12 = 14.84

(iii) Find a and b,
where P(a < X < b) = 0.50
and P(X > b) = 0.25
Solution:
b 
P( X  b)  1  FZ    0.25
  
b
FZ    0.75
  
b

z

b

Mean = 12,  = 4, b = ?
 b  (0.67)(4)  12  14.68

b  a  
P(a  X  b)  FZ    FZ    0.5
     
 a   b
FZ    FZ    .5  0.25
     
a 
  0.67( why)

 a  (4)( 0.67)  12  12  2.68  9.32
Most galaxies take the form of a flatten disc,
with the major part of the light coming from this
very thin fundamental plane. The degree of
flattening differs from galaxy to galaxy. In the
Milky Way Galaxy most gases are concentrated
near the center of the fundamental plane. Let X
denote the perpendicular distance from this
center to a gaseous mass. X is normally
distributed with mean 0 and standard deviation
100 parsecs. (A parsec is equal to
approximately 19.2 trillion miles.)
(a) Sketch a graph of the density for X. Indicate on
this graph the probability that a gaseous mass
is located within 200 parsecs of the center of
the fundamental plane. Find this probability.
(b) Approximately what percentage of the gaseous
masses are located more than 250 parsecs
from the center of the plane?
(c) What distance has the property that 20% of the
gaseous masses are at least this far from the
fundamental plane?
(d) What is the moment generating function for X?
Log-Normal Distribution
The positive random variable Y is

said to have a log-normal
distribution, if logeY is normally
distributed with parameters μ and σ.
That is,
X=logeY is normal
 &  are not mean & standard deviations
of Log-normal random variable.
Section 4.4, Problem 45/ p 146
Let X be normal with mean  and

variance 2. Let G denote the
cumulative distribution for Y = eX
and let F denote the cumulative
distribution for X.
Show that (i) G(y) = F(lny), y>0
(ii) G(y) =F(lny)/y, y>0
(iii) Find density of Y
i) G(y) = P(Y  y) = P(eX  y) = P(X  ln y)
= F(ln y) y>0
G(y)=0, y< 0.
dG( y ) dF (ln y ) d ln y
ii)   F (ln y )
dy dy dy
1
 F (ln y ) , y  0.
y

1 1 
 
ln y  2
dG
 e 2 2
 ,y0
dy 2p   y 
 
 0 otherwise
Hence, the density for Y is given by

 ln y   2
1 2 2
g ( y)  e , y 0
2p  y
 0 other wise .
Problem 46
Let Y denote the diameter in millimeters of

styrofoam pellets used in packing. Assume
that Y has a log-normal distribution with
parameter  = 0.8,  = 0.1.
(i) Find the probability that a randomly
selected pellet has a diameter that
exceeds 2.7 mm
(ii) Find two values of Y such that probability
is approximately 0.95 between these
values?
(i) Find the probability that a randomly
selected pellet has a diameter that
exceeds 2.7 mm

Normal Approximation to
the Binomial Distribution
If X is Binomial with parameters n & p
for large n

X  N np, np(1  p) 
Remark: For practical purposes
approximation is acceptable for values of
n & p such that
[p ≤ 0.5 & np > 5] or [p > 0.5 & n(1-p) > 5]
Correction for continuity
1. P(X = a) = P(a - 0.5 < X < a + 0.5)
2. P(X ≤ b) = P(-∞ < X ≤ b)
= P(-∞ < X ≤ b + 0.5)
= P(X ≤ b + 0.5)
3. P(X ≥ a) = P(a ≤ X < ∞)
= P(a - 0.5 ≤ X < ∞)=P( X ≥ a - 0.5)
4. P(a ≤ X ≤ b) = P(a - 0.5 ≤ X ≤ b + 0.5)
Example
Let X be binomial with n = 20 and
p = 0.3. Use the normal approximation
to each of the following. Compare the
results with values obtained from Table
I of App. A
(i) P[X ≤ 3] (ii) P[3≤ X ≤ 8]
(iii) P[X ≥ 4] (iv) P[X = 4]

(ii) P[3≤ X ≤ 8]
P(3  X  8)
 P(3  0.5  X  8  0.5)
(correction to continuity )
 2.5  6 X  np 8 .5  6 
 P   
 20  .3  .7   
 npq 20 .3 .7 
 P(1.71  Z  1.22)
 FZ (1.22)  FZ (1.71)  0.8888  0.0436
53. Although errors are likely when taking
measurements from photographic images, these
errors are often very small. For sharp images with
negligible distortion, errors in measuring
distances are often no larger then 0.0004 inch.
Assume that the probability of a serious
measurement error is 0.05. A series of 150
independent measurements are made. Let X
denote the number of serious errors made.

(i) In finding the probability of making
at least one serious error, is the
normal approximation
appropriate? If so, approximate
the probability using this method.
(ii) Approximate the probability that
at most three serious errors will
be made.
(i) In finding the probability of making at
least one serious error, is the normal
approximation appropriate? If so,
approximate the probability using this
method.
p = 0.05, n = 150 i.e. p ≤ 0.05
& np = (150)(5)/100 = 7.5 > 5
i.e. Normal approximation is appropriate
Let X* be a normal random variable
with parameters np and (npq).
P( X  1)
 P( X *  1  0.5)(correction to continuity )
 X * np 0 .5  150  0 . 05 
 P  
 npq   
 150 0.05 0 .95 
 7 
 P Z    P( Z  2.67)
 2.67 
 1  FZ (2.67)  1  .0044
(ii) Approximate the probability
that at most three serious
errors will be made.
P( X  3)
 P( X *  3.5)(correction to continuity )
 X *  np 3.5  150  .05 
 P  
 npq   
 150 . 05 .95 
 4 
 P Z    P( Z  1.5)
 2.67 
 FZ (1.5)  .0668
Example
The probability that an electronic

component will fail in less than 1000
hours of continuous use is 0.25. Use
the normal approximation to find the
probability that among 200 such
components fewer than 45 will fail in
less than 1000 hours of continuous
use.

Solution
Let X be the number of components out of

200 which fail in less than 1000 hours of
continuous use.
X has binomial distributions with parameter
n = 200 and p = 0.25
Required probability = P[X < 45] = P[X ≤ 44]
Since X is a discrete r.v. taking the values 0,

1, 2, …, 200.
np= 2000.25 = 50, npq = 37.5
(p ≤ 0.5 & np > 5, so normal
approximation acceptable)
Using correction of continuity.
Required probability
 X * 50 44.5  50 
 P X *  44.5  P  
 37.5 37.5 
 PZ  0.9  0.1841
Example
A Safety engineer feels that 30% of all

industrial accidents in her plant are caused
by failure of employees to follow
instructions. If this figure is correct, find,
approximately, the probability that among 84
industrialized accidents in this plant
anywhere from 20 to 30 (inclusive) will be
due to failure of employees to follow
instructions.

25.2,
Normal Approximation to the
Poisson Distribution
Let X be Poisson with parameter s.
f ( x) 
e  s
 s 
x
, x  0,1,2,3,.......
x!
Then for large values of s, X is
approximately normal with mean s and
variance s.

Example
Let X be a Poisson random
variable with parameter s = 15.
Find P(X  12) from table II of
App. A. Approximate this
probability using normal curve. Be
sure to employ the half-unit
correction.

Solution : P(X  12)  0.268(tabular Value)
P( X  12)
 P( X *  12.5)(correction to continuity )
 X * 15 12.5  15 
 P  
 15 15 
  2.5 
 P Z    P( Z  0.65)
 3.87 
 FZ (0.65)  1  FZ (0.65)
 0.2538 BITS Pilani, Pilani Campus
12 The average number of jets either
arriving at or departing from O’Hare
Airport is one every 40 seconds.
(i) What is the approximate probability
that at least 75 such flights will
occur during a randomly selected
hour?
(ii) What is the probability that fewer
than 100 flights will take place in an
hour? BITS Pilani, Pilani Campus
(ii)  = 1 for unit of 40 sec. But interval
has length one hour i.e. 3600 sec.
Thus s = 3600/40, hence parameter s
= 90
P( X  75)  1  P( X  75)
 1  P( X  74)  1  P( X *  74.5)
 X * 90 74.5  90 
 1  P  
 90 90 
 P( Z  1.635)  FZ (1.635)  0.05105
(ii)  = 1 for unit of 40 sec. But unit is
one hour i.e. 3600 sec. For 3600 sec
parameter s = 90
P( X  100)
 P( X  99)  P( X *  99.5)
 X * 90 9.5 
 P  
 90 90 
 P( Z  1.00)
 FZ (1)  0.8413
BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956
Normal Probability Rule and
Chebyshev’s ineqality
It is sometimes useful to have a quick

way of determining which values of a
random variables are common and
which are considered to be rare.

Normal Probability Rule
Let X be normally distributed with
parameters mean µ and s.d.  . Then
P[- < X-µ <  ]  0.68
P[-2 < X-µ < 2 ]  0.95
P[-3 < X-µ < 3 ]  0.997

The number of Btu’s of petroleum and
petroleum products used per person in the
United States in 1975 was normally
distributed with mean 153 million Btu’s and
standard deviation 25 million Btu’s.
Approximately what percentage of the
population used between 128 and 178 million
Btu’s during that year? Approximately what
percentage of the population used in excess
of 228 million Btu’s?
Solution
Let X is number of Btu’s of petroleum

and petroleum products used per person
in the United States in 1975.
X is N(153, 25)
P[128 < X < 178]
= P[128 – 153 < X- 153 < 178 – 153]
= P[- 25 < X – 153 < 25]
= P[-  < X –  < ]
From Normal Probability law
P[-  < X –  < ]  0.68
Therefore,
P[128 < X < 178]
= P[- 25 < X – 153 < 25]
 0.68

P[X > 228] = P[X – 153 > 228 – 153]
= PX – 153 > 75]
= P[X –  > 3] (as   25
From Normal Probability law
P[- 3 < X –  < 3]  0.997
Therefore,
P[X –  > 3]  (1 - 0.997)/2
 0.0015
Chebyshev’s Inequality
Let X be a random variable with mean 
and standard deviation . then for any
positive number k,
P[ |X - µ| < k] ≥ 1 - 1/k2
P[ |X - µ| ≥ k] ≤ 1/k2
where  is the mean and  is standard
deviation.
Put k = , then P[ |X - µ| ≥  ] ≤ 2/2
Example
Show, by Chebyshev’s inequality, that in
2000 throws with a coin, the probability
that the number of heads lies between 900
and 1100 is at least 19/20. (900 and 1100
not included)
Solution: Let X be the number of heads
(successes) in throwing of a coin, then
X ~ Bin(n,p), where n = 2000, p = 1/2.
E(X) = np = 1000, Var(X) = npq = 500.
To find a lower bound for
P900  X  1000  1100

 P(| X  1000 |  100)
By Chebyshev’s theorem,
(1)
P X    k   1  2
1
k
  1
i.e. P X  1000  500 k  1  2
k
Letting
100
500 k  100, i.e., k 
500
10000
k 
2
 20
500
1 1 19
Therefore, 1  2  1  
k 20 20
P900  X  1100 
19
20
X = number of bulletproof vests out of 405
which fail to meet puncture standards.
X has binomial distribution with n = 405, p =1/6.
μ= 405/6 = 67.5, σ = ((67.5)5/6) = 7.5.

Find the bound on probability if 900 &
1100 are included?
P900  X  1100   P899  X  1101
 P899  1000  X  1000  1101  1000 
 P  X  1000  101
101
k  101  k 
500
1 500
 1 2  1 2
 0.951
k (101)
P X  1000  101  0.951
Example
From a usual pack of 52 cards,
cards are drawn randomly with
replacement till the red card
appears. If X denotes the number
of card drawn, using Chebyshev’s
inequality, find a lower bound for
P[I X - 2 I < 2].

Example
If X is Gamma random variable

with a = 0.05 & b = 100 find an
upper bound on
P([(X-4)(X-6)]  999).

Exercise
How many times should we toss a

balanced coin so that we can assert with
probability at least 0.99 that the proportion
of heads occurs between 0.45 and 0.55?
Solution : Let n be number of trials.
Now on each trial, p = probability of
head = 0.5.
X = number of heads in n trials.
X has binomial distribution with
parameters n and p=0.5, thus =n/2 and
2=n/4.
Proportion of heads =X/n is between
0.45 and 0.55
if and only if 0.45n < X < 0.55n. (*)
Since X has binomial dist. With
p=1/2,
=0.5n.
Thus (*) is equivalent to
|X- | < 0.05n.
The required probability = P(|X-| < 0.05n)
which will be at least 1 –1/k2
where k =k n1/2/2 =0.05n.
This gives k=0.1 n1/2.
Thus if we choose n such that
1 –1/k2 = 1- 100/n > 0.99
then we get what was desired.
Now if n > 10000, then this condition is

satisfied.
This justifies the relative frequency approach
to the probability. The proportion of successes
(i.e. relative freq) can be brought as close to
the actual probability by increasing the number
of trials.
Example
In one out of 6 cases, material for
bulletproof vests fails to meet
puncture standards. If 405
specimens are tested, what does
Chebyshev’s theorem tell us about
the probability of getting at most 30
or more than105 cases that do not
meet puncture standards?
Question 4.5.50,pg.147
For a normal distribution

P[|X-µ| < 3] = 0.997. What value is
associated to this probability via
Chebyshev’s inequality? Are the
results consistent? Which rule gives a
stronger statement in case of a normal
variable?

CH 4 For Display

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

CH 4 For Display

Încărcat de

Drepturi de autor:

Formate disponibile

BITS Pilani

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

Note : F is cdf of continuous rv if and only if 1., 2, 3.

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

(e) Let (l, m) and (d, f) be subintervals of (a, c)

BITS Pilani, Pilani Campus

4. m.g.f. = Moment Generating Function

BITS Pilani, Pilani Campus

Let X denote the length in minutes of a long

BITS Pilani, Pilani Campus

If a pair of coils were placed around a

BITS Pilani, Pilani Campus

q is uniformly distributed over the interval [0, 2p].

a) Find the density for q.

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

Multiply and divide by 2, we get

BITS Pilani, Pilani Campus

Definition : The function

BITS Pilani, Pilani Campus

by integrating by parts, we have

BITS Pilani, Pilani Campus

Convert to polar coordinates and see it is π.

BITS Pilani, Pilani Campus

is said to have a Gamma Distribution with

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

Let X be a Gamma r.v. with parameters

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

 a(a  1)b  (ab)  ab

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

For the above function

mX(t)=(1- bt)-1 , t < 1/b

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

The distribution function F for W is given by

BITS Pilani, Pilani Campus

This is exactly density for an exponential

A random variable X with density function

is said to have a Gamma Distribution with

BITS Pilani, Pilani Campus

If a random variable X has a gamma

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus

5 1.61 2.67 4.35

If F is CDF for Chi square random variable

10 16.0 18.3 20.5

BITS Pilani, Pilani Campus

Sol:  is random var iable with

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus