L3: Special distributions

Parameter – a special constant, when allocated

certain value will determine the exact probability
distribution of a r.v.

Bernoulli distribution
i) Random experiment with 1 trial.
ii) The outcome of which can be classified in
but one of two mutually exclusive and
exhaustive ways, say, success or failure
(e.g. female or male, life or death,
nondefective or defective).
iii) P (success)  p
iv) P (failure)  1  p  q
v) A r.v. X is the occurrence of success in 1
trial.

Notation: X ~ B (1, p) . Then the p.d.f. of X is

 p x (1  p )1 x ; x  0,1
f ( x)  
0 ; otherwise

1
Example:
1) Toss a fair of coin.
Success = head appear

2) An observation at a junction.
Success = an accident occurs

2
Binomial distribution
i) random experiment with n identical trials.
ii) the outcome of each trial can be classified in
but one of two mutually exclusive and
exhaustive ways, say, success or failure.
iii) P (success)  p
P (failure)  1  p  q
iv) the probability of success is the same from
one trial to another.
v) the trials are independent.
vi) The r.v. X is the total number of successes
in n (Bernoulli) trials.

Notation: X ~ B (n, p ) . The p.d.f. of X is

 n  x n x
  p (1  p ) ; x  0,1,, n
f ( x)   x 
0
 ; otherwise
where
n n n!
 x   Cx  x !(n  x)!
 
n!  n(n  1)(n  1)(2)(1)

3
Example:
1) Toss a fair coin 10 times.
success = obtain head
P ( H )  1/ 2 for each toss.

2) 5 balls are selected at random with

replacement from a box that contains 3
white balls and 5 red balls.
success = red balls are selected.
P (red ball)=5/8 on each trial.

3) 6 observations at a junction.
success = accident and P(accident) is the
same for each observation.

4
Exercise: A student is given a quiz which consists
of 10 questions true false. Assuming that the
student is not ready and answers the questions by
guessing,

i) what is the probability that

a) exactly 5 questions are correctly
answered?
b) he answered more than 2 questions but 
4 correctly?
c) he will pass if  7 questions need to be
answered correctly to pass?

ii) find the number of questions that is expected

to be answered correctly.

5
Refer: New Cambridge table
Binomial table: page 4 – 23

Table gives:
F (r )  P( X  r )
r
  p x q n x
x 0

X ~ B (n  10, p  0.5) p.9

6
Example: Given m.g.f. of X as
mX (t )  (0.3 et  0.7 )6 .
Find the mean, variance and p.d.f. of X .

7
Example: A box of N marbles consists of K glass
marbles and N  K stone marbles of the same size.
A marble is selected at random from the box, its
type is noted. Then the marble is returned into the
box. n trials are performed. If X  the number of
glass marble selected, find the distribution of X .

8
Example: Refer to previous example. Let us take a
random sample of n marbles. The number of each
type of marbles are observed and the marbles are
not returned into the box. If out of n selected
marbles, X are glass marbles and n  X are stone
marbles. Find the distribution of X .

N
We can sample it with   ways.
n 

1
Thus P (each sample)  .
N
n 
 

For each sample that contains n marbles, the

number of ways X glass marble can be selected
K 
without concerning the order is   and the
X
number of ways n  X stone marbles selected is
N K
 n  X .
 

9
Since this process happens at the same time, there
 K  N  K 
are    ways to choose X glass marbles
 X  n  X 
and n  X stone marbles altogether.

Thus, the probability of choosing X glass marbles

and n  X stone marbles altogether or the
probability there are X glass marbles in the sample
of n marbles is

 K  N  K 
 x  n  x 
P ( X  x)    , x ?
N
n 
 

10
Hypergeometric distribution

i) Assume a population consists of n objects

which can be classified into 2 types, and let
n1  the number of objects from the first
type.

ii) A sampel of size r  n is taken from the

population without replacement.

iii) X  the number of objects of the first type

selected.
X  0,1, 2,, r

  n1  n2 
   
 
x r  x  ; x  0,1, 2,, r

iv) P ( X  x)   n
 r 
  
 0 ; otherwise

where
x  r, x  n1 , r  x  n 2

It can be shown that

rn
  E ( X )  1 and
n
r n n (n  r )
 2  Var ( X )  12 2
n (n  1)
11
If sampling was performed with replacement, then
X ~ Bin(n, p  nn1 ) .

The mean for both situations are the same.

Assuming the population size is large, n  , then

nr
 1, therefore the variance of the
n 1
hypergeometric distribution approaches binomial
distribution’.

We can show that when n  , and n1   with

p  nn1 , then the p.d.f. of the hypergeometric
distribution converges to the p.d.f. of the binomial
r.v..

Example: In a small pond there are 50 fish, 10 of

which have been tagged. If a fisherman’s catch
consists of 7 fish, selected at random and without
replacement, and X denotes the number of tagged
fish, what is the probability that exactly 2 tagged
fish are caught?

12
Geometric distribution

i) experiment consists of Bernoulli trials which

is performed until a certain number, say r ,
of successes occurs.
ii) each trial has probability of success p ,
0  p  1.
iii) these Bernoulli trials are independent.
iv) X is the number failures before the first
(r  1) success.

The p.d.f. of X is
(1  p ) x p ; x  0,1,
f ( x)  
 0 ; otherwise

We say that X has a geometric distribution. Note

that the first success occurs on the ( x  1) st trial.

Recall: Geometric series


a

k 0
a r k

1 r
; | r | 1

13
If we let Y  X  1, the trials on which the first
success is observed, then

Example: Some biology students were checking

the eye colour for a large number of fruit files. For
the individual fly, suppose that the probability of
white eyes is ¼ and the probability of red eyes is
¾, and that we may treat these flies as independent
Bernoulli trials.

P (3 red-eyed flies are observed before the first white-eyed fly)

=

P (at most four flies have to be checked to observe

a white-eye fly)
=

14
Negative Binomial distribution

i) A sequence of independent Bernoulli trials

until exactly r successes occur.
ii) each trial has probability of success p,
0  p  1.
iii) X is the number failures before the
occurrence of r th success.

Then ( x  r ) denotes the number of trials necessary

to produce exactly r success and x failure, with the
r th success occurring at the ( x  r ) th trial.

The p.d.f. of X is

 x  r  1 r
f ( x)    p (1  p ) x ; x  0,1, 2,
 r 1 
 x  r  1 r
  p (1  p ) x
; x  0,1, 2,
  x 

 0 ; otherwise

Example: Suppose that the Biology students in the

previous example check the eye colour of fruit flies
until the third white-eyed fruit fly is observed.

15
P(of observing 10 fruit flies with red eyes before
the third white-eyed fruit fly is observed) =

P(at most 12 flies have to be checked to find the

third white-eyed fly) =

The reason for calling this negative binomial

distribution:
Let h( w)  (1  w)  n . Using Maclaurin’s series
expansion, we have


h k (0) k
(1  w) n
 w
k 0 k!
 n  k  1 k

  w , | w | 1
k 0  k 
known as negative binomial expansion

16
Example: Given the m.g.f. of a discrete r.v.
4
 0.45 
mX (t )   t 
 1  0.55 e 
What is the p.d.f. of X ? E ( X ) ? Var ( X ) ?

17
Poisson distribution
Let the number of changes (events) that occur in a
given continuous interval be counted. A r.v. X is
the number of changes (events) in this unit interval.
X is said to be a Poisson r.v.. The events that are
observed in the interval are considered as ‘success’.

Notation: X ~ P ( )

The p.d.f. of X is
 e   x
 ; x  0,1, 2,
f ( x)   x !
 0 ; otherwise


where   mean number of events that occur in the

interval.

Example:
1) the number of incoming calls in a minute.
2) the number of typo error on each page.
3) the number of micro-organism in 1 ml water.
4) the number of people that arrive at the post
office in 1 hour.

18
Example: Assume that the accidents that occur at
an iron factory follow a Poisson distribution of
mean ½ per week.

a) Find the probability that

i)  2 accidents occur in 1 week,
ii) no accident happens in 6 weeks,
iii)  4 accidents occur in 6 weeks.

b) Find the expected number of weeks in 6 weeks

selected at random where no accident occurs.

19
Poisson distribution as an approximation of
Binomial distribution

A Poisson distribution can be used as an

approximation to binomial distribution when the n
is large and p is small.

This approximation will help us in terms of

computation since there are 2 parameters involved
in binomial distribution where there is only 1
parameter in Poisson distribution.

20
Thus,
n x n x e np (np ) x
f ( x)    p (1  p ) approaches

 x x!

if n   and p  0 (small).

Notes: Also, for large n , Binomial table cannot be

used. Therefore, Poisson approximation is used.

Example: X ~ B(300,0.02)

21