Scribd
Upload
75 views

The Poisson Distribution: History

The Poisson distribution models the number of events occurring in a fixed interval of time or space if these events happen with a known average rate and independently of the time since the last event. It was originally derived by Simeon Poisson in the early 1800s as an approximation to the binomial distribution for rare events. The Poisson distribution has expected value and variance both equal to the rate parameter lambda. It is closely related to the exponential distribution and models processes like queues where arrivals follow a Poisson process.

Uploaded by

Ayush Jain
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as RTF, PDF, TXT or read online on Scribd
75 views

The Poisson Distribution: History

The Poisson distribution models the number of events occurring in a fixed interval of time or space if these events happen with a known average rate and independently of the time since the last event. It was originally derived by Simeon Poisson in the early 1800s as an approximation to the binomial distribution for rare events. The Poisson distribution has expected value and variance both equal to the rate parameter lambda. It is closely related to the exponential distribution and models processes like queues where arrivals follow a Poisson process.

Uploaded by

Ayush Jain
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as RTF, PDF, TXT or read online on Scribd
You are on page 1/ 7

The Poisson Distribution

History
The Poisson distribution was originally derived by Simeon Denis Poisson (1781-1840) in 18 8! "e was a #ren$h mathemati$ian% geometer and &hysi$ist% and 'nown as the (irst &erson to derive the underlying (ormula (or the Poisson distribution! )t was more or less a by-&rodu$t o( Poisson*s dis$ussion o( +aw o( +arge ,umbers in his boo' Sur la probability des judgments &ublished in18 7! )n the $ha&ter the law o( large numbers% he said that the &robability o( the number o( ha&&enings o( an event - in trials will (all within $ertain assigned limits! .sing the binomial% he (ound that (or large values o( n it would be easier to avoid the /ui$'ly growing (a$torial i( n is allowed to in$rease to in(inity and & to a&&roa$h 0ero! This $on$e&t% whi$h is today $alled the +aw o( Small ,umbers% was truly novel! "is observation% however% was not noti$ed by the &ubli$ until 1818% when it was summari0ed by +adislaus von 2ort'iewi$0 ! +adislaus von 2ort'iewi$0 (1838 - 11 1) was a 4erman &ro(essor in .niversity o( 2erlin! "e was the (irst &erson to &oint out the im&ortan$e o( Poisson distribution% and he &ublished a monogra&h entitled The Law of Small Numbers that illustrated the im&ortan$e o( Poisson5s (ormula! The law o( small numbers sim&ly &oints out that the $hara$teristi$s o( a small sam&le $annot be said to be true (or a &o&ulation! #inally% Poisson*s theory was trans(ormed (rom Poisson*s 6limit7 to Poison*s 6distribution7% as (ollows!

, ';0%1%<= +et; n&% and also assume;n& thstatyswhen$o n a&&roa$hes in(inity and & a&&roa$hes 0ero%
8ssume 9 : 2(n%&) % then the &d( o( 9 is% (

) =

(1 )

& $an>nbe!Thenwrittenweasare going to rewrite the &d(%


!

stant! Thus%

lim

lim

! lim

lim

2(

=
+ 1)

lim

lim

lim

lim 1

(1) 2e(ore $ontinuing on% we need to introdu$e two lemmas! Lemma 1 #rom the de(inition o( the number e% we get that that ? ; -n>% we $an rewrite e/uation as% the ! 8ssume

= lim

lim

= lim

1+

)=

! (<)

Lemma 2 The value o( lim

= lim

2y &lugging (<) and ( ) ba$' into (1)% we $an see that%

;1! ( )

!1

lim

= ! .

Definition
@hi$h im&lies that the &d(
= =
!

. ';0%1%<!!!% A0

where ' is the actual number o( observationsistheo( the event being tra$'ed during the given interval and expected number o( observations during the given interval! This derivation shows why the Poisson distribution $an be regarded as an a&&ro?imation to the binomial% whi$h% be$ause o( (a$torials involved% $an be di((i$ult to $al$ulate due to $ertain values o( ' and n! #rom this (ormula% we $an also easily derive $d(B

$d( P(9 C; ') ;

% ';0%1%<!!!% A0

Expected value and Variance The Poisson distribution is dedu$ted (rom a&&ro?imating a binomial random variable with &arameters (n%&) when n is large and & is small% and ; n&! Sin$e the binomial random variable has an e?&e$ted value n& ; and varian$e n&(1-&) ; (1-&) % when & 0% it seems reasonable that a Poisson Dariable would have both its e?&e$ted value and varian$e e/ual to as well! To (ind the e?&e$ted value% we $an $om&ute the (ollowingB
EE X F =

ie
i =0

iG

i i1 = e
= e

i1

+etting H ; i1

i=1

i G = j G

(i 1)G
i=1

= e

j =0

,e?t% we $an determine its varian$e! #irst% re$all that DarE9F ; -E9 F I (-E9F) ! So now we Hust < need to (ind -E9 F!

<

<

E[X ] =
2

i2e

2=
e j

i=0

i!
i<

1=

i1

= e

=1

iG

= ( +1)

j =0

jG

+ e

j =0

jG

ii 1
i

( j

= 1

Sin$e we have shown that -E9F ; % we obtain that

+1) j
j =0

jG

Var[X] = E[ X 2 ] ( E[ X])2 = 2 + 2 =
Thus% both the e?&e$ted value and varian$e o( a Poisson random variable are e/ual to as e?&e$ted!

To see how the Poisson a$ts as an a&&ro?imation to the binomial% we $an $om&are the two distributions &i$torially!

Poisson PJ# with 1 C; C; 1

2inomial PJ#% (or $om&arison

The Poisson is a$$urate (or high values o( n with small values o( p! Pg! <77 o( +arsen and Jar? has a $om&arison o( &robabilities $al$ulated with ea$h model!

Relationships with other distributions


The Poisson is related to many other distributions% not Hust the binomial! )t is &erha&s most $losely related to the e?&onential and the gamma! The e?&onential distribution is more o(ten a&&lied in business models% su$h as in the &roblem o( /ueuing system o&timi0ation! )( the arrival o( $ustomers has a Poisson distribution with a mean o( % the waiting time (or the $ustomers in a /ueue would (ollow an e?&onential distribution with a mean time o(

! To de(ine the e?&onentialB su&&ose a series o( events satis(ying the Poisson model are o$$urring at the rate o( &er unit time! +et the random variable K denote the interval between $onse$utive evens! Then K has the e?&onential distribution%

To draw its gra&h% we let S ; 9 denotes the number o( o$$urren$es in a unit o( time and 9 has a Poisson distribution% and T;K denotes the interval between $onse$utive o$$urren$es% namely% the waiting time% and the waiting time K has an e?&onential distribution! The gra&hi$al relationshi& between the o$$urren$e S and the waiting time T is shown as belowB

Than you for evaluatin!

"#L easy#onverter Des top


This @ord do$ument was $onverted (rom PD# with an evaluation version o( 2L+ easyLonverter Des'to& so(tware that only converts the first $ pa!es o( your PD#! 8$tivate your so(tware (or less than M<0
htt&B>>www!&d(online!$om>easy$onverter>

LTN+O Lli$' on the lin' below to &ur$hase

You might also like