Lect 5 Geometric Distribution

You might also like

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 20

Lecture 5

5
AAOC C111: PROBABILITY &
STATISTICS
BITS-PILANI HYDERABAD CAMPUS
Presented by
Dr. M.S. Radhakrishnan
Email: msr@bits-hyderabad.ac.in
Lecture 5
Geometric Distribution

Article 3.4
Text Book: J. SUSAN MILTON and
JESSE C. ARNOLD, Introduction to
Probability and Statistics, Tata McGraw-
Hill Edition, Fourth Reprint 2008.
5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 2
In this lecture we look at

• Geometric Distribution
• Mean and Variance of a geometric
density
• Moment Generating Function
(mgf) of geometric density
• Some simple examples
5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 3
5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 4
Geometric Distribution
Suppose we have a random experiment E that
consists of only 2 outcomes, say “success”
and “failure”. Let the probability of getting a
success be p and so the probability of getting
a failure is 1 – p = q. We make the further
assumption that p, the probability of getting a
success remains the same for all independent
repetitions (= trials) of the experiment E. That
is the experiment “does not learn.”
5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS-Pilani 5
We repeat the experiment till the first
success occurs.

Let X be number of trials needed to get the


1st success.
Clearly X is a discrete r.v. that can take
values 1, 2, 3, …
We say X has a geometric distribution
with parameter p.

5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS-Pilani 6


The density of X is given by
g (x; p) = P( X = x)
= P(first x – 1 trials were failures and
the xth trial was a success)
= qx-1 p, x = 1, 2, …

p
 g ( x; p )  
x 1
q x 1
p 
1 q
 1

Since g (x; p)  0 for all x, the above is a


valid density.
5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS-Pilani 7
cdf of the Geometric distribution.
x

G(x; p) = P (X  x)   g ( x; p)
k 1

x 1 qx
 q k 1
p p  1 qx
k 1
1 q

Mean of the Geometric Distribution


 
1
 = E(X)   x  g ( x; p )   x  q x 1
p 
x 1 x 1 p

5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS-Pilani 8


Explanation:

1
  qx |q|<1
1  q x 0

Differentiating both sides w.r.t. q, we get


2
 1  

 
x 1
  x q
 1 q  x 1

Multiplying both sides by p, we get the


result (noting that 1- q = p).
5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 9
Variance of the Geometric
Distribution
E(X2) = E[X(X-1)+X] = E[X(X-1)] + E(X)
 
E[X(X-1)]   x( x  1)  g ( x; p)   x( x  1)q x1 p
x 1 x 1


2 2q
 pq  x( x  1)q x2
 pq  2
x 2 (1  q)3 p

5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 10


Hence the variance of X is

  Var ( X )  E ( X )  
2 2 2

2q 1 1 2q  p  1 2q  q
 2  2  
p p p p2 p2

q
 2
p

5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 11


MGF of the Geometric Distribution
mgf of X is
 

MX(t) = E(etX)   e P( X  x)
tx
  etx  q x1 p
x 1 x 1

  qe 
x 1 pet
 pe t t

x 1 1  qet
for all t for which qet < 1
i.e. for all t for which t < - ln q.

5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS-Pilani 12


Mean of the Geometric distribution
(using mgf)
pet
mgf of X is M X (t )  , t   ln q
1  qe t

M  (t )  p
1  qe  e  qe
t t 2t

pe t

1  qe  1  qe 
X 2 2
t t

Hence mean of X is
p 1
  M X (0)  
1  q 
2
p
5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 13
Variance of the Geometric
Distribution

 
2
1  qe t
et  2qe2t (1  qet )
M X (t )  p
1  qe 
4
t

p
 
1  qet et  2qe2t
p
1  qe2t
1  qe  1  qe 
3 3
t t

5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 14


1 q
Hence 2  E ( X 2
)  M X (0) p
1  q 
3

1 q
 2
p

Thus the variance of X is


1 q 1
  Var ( X )  2  
2 2
 2  2
p p
q
 2
p
5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS-Pilani 15
Memoryless Property of the
Geometric Distribution
Let s, t be two positive integers. Then
P(X > s+t | X > s) = P (X > t)

Proof
q t.
We note that P( X > t) =

5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS-Pilani 16


Now
P[( X  s  t )  ( X  s)]
P(X > s+t | X > s) 
P( X  s )

Note that the event


P( X  s  t )
 (X > s + t) is a subset of
P( X  s ) the event (X > s)
s t
q
 s q t
= P( X > t)
q
Q.E.D.
5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 17
Exercise 24 Section 3.4 Page 86
The probability that a wildcat well will be
productive is 1/13. Assume that a group is
drilling wells in various parts of the country
so that the status of one well has no bearing
on that of any other. Let X denote the
number of wells drilled to obtain the first
strike.

5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 18


(a) Verify that X is geometric, and identify
the value of the parameter.
(b) What is the exact expression for the
density for X?
(c) What is the exact expression for the mgf
for X?
(d) What are the numerical values of
E[X], E[x2], 2, and ?
(e) Find P[ X  2].
5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 19
Exercise 35 Section 3.4 Page 88

Let the density for X be given by


f (x) = c e-x, x = 1, 2, 3,…
(a) Find the value of c that makes this a
density.
(b) Find the mgf for X.
(c) Use the mgf to find E[X].

5-Aug-19 Prepared by Dr. M.S. Radhakrishnan, BITS, Pilani (Rajasthan) 20

You might also like