Scribd
Upload
70 views6 pages

Noncentral Hypergeometric Distribution: J S M J M P

The document describes the noncentral hypergeometric distribution and its properties. It discusses both the univariate and multivariate cases. For the univariate case, it provides the formula for the probability mass function and defines key parameters. It then derives the expectation and variance in terms of these parameters. For the multivariate case, it similarly defines the probability mass function and derives formulas for the conditional expectation and covariance between variables.

Uploaded by

juntujuntu
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
70 views6 pages

Noncentral Hypergeometric Distribution: J S M J M P

The document describes the noncentral hypergeometric distribution and its properties. It discusses both the univariate and multivariate cases. For the univariate case, it provides the formula for the probability mass function and defines key parameters. It then derives the expectation and variance in terms of these parameters. For the multivariate case, it similarly defines the probability mass function and derives formulas for the conditional expectation and covariance between variables.

Uploaded by

juntujuntu
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

Noncentral hypergeometric distribution

(i) Univariate noncentral hypergeometric distribution There are several ways to derive the univariate noncentral hypergeometric distribution. Source 1 The noncentral hypergeometric distribution with odd ratio is an

exponentially weighted version of the central hypergeometric distribution. Let Y has a noncentral hypergeometric distribution. Then,

m1 m 2 y s y 1 P (Y = y | ) = P0 ( )
where

m1 m 2 y y s y 1 = b m1 m 2 j s j j=a 1

P0 ( ) =

m1 m 2 j j=a s1
b

and

a = max

(0 , s 1

m 2 ) y min

(m 1 , s 1 ) =

b .

The distribution arises in the exponentially weighted sampling

m scheme in which each of the m 1


proportionally to
y

possible samples is weighted

. We denote

Y ~ H (m , s ;

or Y ~ H (s , m ;
1

),

where

m = (m 1 , m 2

and

s = (s 1 , s 2 ) . Note that

H (m , s ;1 ) = H (m , s )
is the central hypergeometric distribution. Source 2 Let Y ~ B (m 1 ,

),

X ~ B (m

and

1 1
2

.
2

Then, the conditional distribution of Y given X + Y = s 1 is the noncentral hypergeometric distribution

H (m , s ; ) .

The expectation and variance of obtained as follows. Let

Y ~ H (m , s ;

can be

= log (

) . The distribution of

is

y f ( y , ) exp log ( ) P 0
= exp = exp

{y log ( ) log [P0 ( )]} {y log [P0 ( )]}

It is the standard form of exponential family. The cumulant generating function

k (t ) = log P 0 e t + E (Y
where

[ (

)] log [P (e )] P (e )

0 t +

)=

'

(0 ) =

P0 e

( )
j

t=0

P1 ( P 0 (

) )

P0 e t + t t=0

b t + e j=a =

m1 m 2 j s1 t

j t=0

= P1 (
Similarly,

b m2 j m1 = j e t + j s 1 j=a b m1 m 2 = je j j s j j=a 1 b m1 m 2 = j j j s j j=a 1

j t=0

P ( ) P1 ( ) Var (Y ) = k ' ' (0 ) = 2 P0 ( ) P0 ( )


where

2 P0 e t + P2 ( ) = = 2 t t=0

j=a

m1 m 2 j 2 j j s1

j .

(i) Multivariate noncentral hypergeometric distribution Let


3

Y = (Y1 , Y2 ,K, Yk ) ~ M (m1 , 1 ), X = ( X 1 , X 2 ,K, X k ) ~ M (m2 , 2 )


where

1 = ( 11 , 12 , K , 1 k ) and

= (

21

22

,K ,

2k

).

Then, the conditional distribution of Y given X + Y = s

is

m 1 m 2 y1 y 2 yk L k y s y 1 2 P (Y = y | X + Y = s ) = m1 m 2 j1 j 2 jk j s j 1 2 L k j
where

1 j 1 j

2 j 2 j

The conditional expectation and variance of Y can be derived as follows. The conditional distribution of Y is

(y

| s ,

)=

P (Y = y | s , log

k exp y j j =1
where

[P 0 ( )]

= log (

) and

m1 m 2 j = a j s1
b

P0 ( ) = P0 ( 1 , 2 , K , k ) =

j1 1 L kj k j .

Since it is the standard form of exponential family, the cumulant generating function is

k (t ) = k (t1 , t 2 , K , t k ) = log P0 e t 1 + 1 , e t 2 + 2 , K , e t k + k
Then,

[ (

)] log [P (e
0

, e 2 , K , e k

)]

E (Y l

)=

P0 e t 1 + 1 , e t 2 + 2 , K , e t k + k tl tl = 0 = P0 e 1 , e 2 , K , e k =
where

k (t ) tl t=0

P1 , l (

) P 0 ( )

P0 e t 1 + 1 , e t 2 + 2 , K , e t k + k t l tl = 0 j = m 1 m 2 (t 1 + 1 ) j 1 L e (t k + k ) j k j s j e tl t=0

= j = =

m 1 m 2 (t k + k ) j k (t 1 + 1 ) j 1 L j e e j s j l t=0 m 1 m 2 k jk 1 j1 L j e e l j s j

= P1 , l (

m 1 m 2 jk j1 j2 j L k j s j l 1 2

Similarly,

2 k (t ) Cov (Y l , Y m ) = tltm t=0 =


where

P2 , lm ( P0 (

P ( ) P1 , m ( ) 1,l ( ) ( ) P P 0 0

P2 , lm ( ) =

m1 m 2 jk j1 j2 j j L 1 2 l m k j s j

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy