Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
22 views

Y e y Y P: (A) Poisson Distribution and Poisson Process

1. The document discusses the Poisson distribution and Poisson process. It provides the probability mass function and cumulant generating function of the Poisson distribution. 2. It defines three equivalent definitions of a Poisson process. A Poisson process has independent and stationary increments with the number of events in any time interval following a Poisson distribution. 3. The Poisson log-likelihood function for a sample of observations from a Poisson distribution is presented. The deviance function, which is related to Pearson's statistic, is used to compare maximum likelihood models.

Uploaded by

juntujuntu
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
22 views

Y e y Y P: (A) Poisson Distribution and Poisson Process

1. The document discusses the Poisson distribution and Poisson process. It provides the probability mass function and cumulant generating function of the Poisson distribution. 2. It defines three equivalent definitions of a Poisson process. A Poisson process has independent and stationary increments with the number of events in any time interval following a Poisson distribution. 3. The Poisson log-likelihood function for a sample of observations from a Poisson distribution is presented. The deviance function, which is related to Pearson's statistic, is used to compare maximum likelihood models.

Uploaded by

juntujuntu
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

1

4.2 Likelihood Functions



(a) Poisson distribution and Poisson process

Poisson distribution:

The distribution of a Poisson random variable Y is
( )
! y
e
y Y P
y


= =
.
The cumulant generating function is
( ) ( ) 1 =
t
Y
e t k
.
Thus,
( )
( )
| | = =
(

=
=
=
0
0
t
t
t
Y
e
t
t k
Y E

and
( )
( )
| | = =
(

=
=
=
0
0
2
2
t
t
t
Y
e
t
t k
Y Var
.

Important Result:
Any random variable X whose cumulants are
( ) n O
, where n is
some quantity tending to infinity, has the limiting property

( )
( )
|
.
|

\
|
+

=

2
1
2
1
2
1 , 0 ~ n O N
X Var
X
k
X
p

,
2
where
( ) X E =
. In particular, for the Poisson random variable Y,
as
( ) = Y Var
,
( )
|
.
|

\
|
+

2
1
2
1
1 , 0 ~ n O N
Y
p




Important Properties:
For large

, the variance-stabilizing transform is


2
1
Y in the sense
that
4
1
,
2
1
2
1
2
1

|
.
|

\
|

|
.
|

\
|
Y Var Y E
.

The other transformation which approximates standard Normal is
( )
( )

= +
+
=

0 ,
6
2
0 ,
6
3 3
2
1
2
1
2
1
3
1
6
1
2
1
Y
Y Y Y
Y g



Poisson process:

There are 3 equivalent definitions for Poisson process. They are:
Definition 1:
Suppose events are occurring successively in time, with the intervals
between successive events independently and identically
distributed according to an exponential density
3
( )
t
e t f


=
.
Denote the number of events during
| | t , 0
by
( ) t N
. Then, the
stochastic process
( ) { } 0 ; t t N
is called a Poisson process with
mean rate (or intensity)

.
Definition 2:
An integer valued process
( ) { } 0 ; t t N
is a Poisson process with
mean rate

if
(a)
( ) { } 0 ; t t N
has stationary independent increments.
(b) For time points s and t,
t s <
, the number
( ) ( ) s N t N
of
counts in the interval
| | t s ,
has a Poisson distribution with mean
( ) s t
:
( ) ( ) ( )
( )
( ) | |
! k
s t e
k s N t N P
k
s t

= =


Definition 3:
As in Definition 1, let
( ) t N
denote the number of events that have
occurred during
| | t , 0
. Suppose that whatever the number of events
during
| | t , 0
,
(a) the probability of an event during
| | h t t + ,
is
( ) h o h +
.
(b) the probability of more than one event during
| | h t t + ,
is
4
( ) h o
.
Then, the stochastic process
( ) { } 0 ; t t N
is called a Poisson
process with mean rate

.

(b) The Poisson log-likelihood function

The Poisson log-likelihood function for
n
Y Y Y , , ,
2 1
K
is
( ) ( ) ( )

=

n
i
i i i n
y l
1
2 1
log , , , K
,
where
( )
i i
Y E =
. The deviance function is
( ) ( ) ( )
( )
( )

= =
=

(
(

|
|
.
|

\
|
=
(
(


|
|
.
|

\
|
=
=
n
i
i i
n
i
i
i
i
n
i
i i
i
i
i
n n n n
y
y
y
y
y
y
y y l y y D
1 1
1
1 1 1 1
2

log 2

log 2
, , 2 , , 2 , , , , ,

K K K K

where
i

is the estimate of
( )
i i
Y E =


Note:
If a constant term is included in the model, it can be shown that

( ) 0
1
=

=
n
i
i i
y
.
Thus, the deviance function can be reduced to
5
( )

= (
(

|
|
.
|

\
|
=
n
i
i
i
i n n
y
y y y D
1
1 1

log 2 , , , , ,

K K
.
Note:
The deviance function is closely related to Pearsons statistic. Since
( )
i i i i
i
i
i
i
i i
y
y y


+ = + = =

1 1

,
then
( ) ( ) ( )
( ) ( ) | |
( )
2


2

2 2

2
1
1 log 1
1 log 1

log
2
2
3
2
2
2
i i
i
i
i
i
i
i
i i
i
i
i i i
i i i i
i i i i i i i
i
i
i
y
y
y

|
|
.
|

\
|
+ =
(

|
|
.
|

\
|
+ + =
(

|
|
.
|

\
|
+ + =
+ + =
+ + =
|
|
.
|

\
|
L
L
L

where
( ) L + = +
2
1 log
2
i
i i


. Thus,
6
( ) ( )
( )
statistic s Pearson'

log 2 , , , , ,
1
2
1
1 1

(
(


|
|
.
|

\
|
=

=
=
n
i
i
i i
n
i
i i
i
i
i n n
y
y
y
y y y D

K K

You might also like