ADDITIVE PARTITION FUNCTIONS AND A CLASS OF STATISTICAL HYPOTHESES

456 102 2MB

English Pages 67

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

ADDITIVE PARTITION FUNCTIONS AND A CLASS OF STATISTICAL HYPOTHESES

Citation preview

INFORMATION TO USERS

This material was produced from a microfilm copy of the original docum ent. While the most advanced technological means to photograph and reproduce this document have been used, the quality is heavily dependent upon the quality of the original submitted. The following explanation of techniques is provided to help you understand markings or patterns which may appear on this reproduction. 1. The sign or "target" for pages apparently lacking from the document photographed is "Missing Page(s)". If it was possible to obtain the missing page(s) or section, they are spliced into the film along with adjacent pages. This may have necessitated cutting thru an image and duplicating adjacent pages to insure you com plete continuity. 2. When an image on the film is obliterated with a large round black mark, it is an indication th at the photographer suspected th at the copy may have moved during exposure and thus cause a blurred image. You will find a good image of the page in the adjacent frame. 3. When a map, drawing or chart, etc., was part of the material being photographed the photographer followed a definite method in "sectioning" the material. It is customary to begin photoing at the upper left hand corner of a large sheet and to continue photoing from left to right in equal sections with a small overlap. If necessary, sectioning is continued again — beginning below the first row and continuing on until complete. 4. The majority of users indicate th a t the textual content is of greatest value, however, a somewhat higher quality reproduction could be made from "photographs" if essential to the understanding of the dissertation. Silver prints of "photographs" may be ordered a t additional charge by writing the Order Department, giving the catalog number, title, author and specific pages you wish reproduced. 5. PLEASE NOTE: Some pages may have indistinct print. Filmed as received.

Xerox University Microfilms 300 North Z eeb Road Ann Arbor, Michigan 48106

ILD3907 ; .G7 W o lfo w itz , Jacob , 1 9 1 0 1942 A d d i t i v e p a r t i t i o n f u n c t i o n s and a ; .V.'7 c la s s o f s t a t is t ic a l h y p o th e s e s ... cilew Y o r k 3 1 9 4 2 . 6 2 , d a ty p ew r itte n l e a v e s . 29cm. T h e s i s ( P h . D . ) - New Y o r k u n i v e r s i t y , G raduate s c h o o l , 1 9 4 2 . " R e f e r e n c e s " : p . e6 3 a A 84750

SbeffList

Xerox University Microfilms,

Ann Arbor, Michigan 48106

T H IS D IS S E R T A T IO N HAS BEEN M IC R O F IL M E D E X A C T L Y A S R E C E IV E D .

ADDITIVE PARTITION FUNCTIONS AND A CLASS OF STATISTICAL HYPOTHESES by / l

W o lfo v /itz 1942.

d i s s e r t a t i o n i n t h e d e p a r tm e n t o f m a t h e m a t ic s s u b m it t e d t o t h e f a c u l t y o f t h e G r a d u a te S c h o o l o f A r t s and S c i e n c e s i n p a r t i a l f u l f i l l m e n t o f the r e q u i r e m e n t s f o r t h e d e g r e e o f D octor o f P h ilo s o p h y .

a

ADDITIVE PARTITION FUNCTIONS AND A CLASS OF STATISTICAL HYPOTHESES by J . W olfow itz 1 . I n tr o d u c tio n The purpose o f th e f i r s t p a r t o f t h i s p ap er i s to prove s e v e r a l theorem s ab o u t a c l a s s o f f u n c tio n s o f p a r t i t i o n s which a r e a d d i tiv e i n s t r u c t u r e and s u b je c t t o ^ l d r e s t r i c t i o n s .

These theorem s may be

re g a rd e d a s c o n tr ib u tio n s t o th e th e o ry o f num bers, b u t i f one makes c e r t a i n assig n m e n ts o f p r o b a b i l i t i e s to th e p a r t i t i o n s th e theorem s may be e x p re s s e d a s s ta te m e n ts a b o u t asy m p to tic d i s t r i b u t i o n s .

I t is

i n t h i s l a t t e r , p r o b a b ilis tic lan g u ag e t h a t we s h a l l c a rry out th e p r o o f s , f o r s e v e r a l re a s o n s .

The d is c u s s io n v a i l be more co n cise and

c e r t a i n c irc u m lo c u tio n s w i l l be a v o id e d .

The theorem s have s t a t i s t i c a l

a p p l i c a t i o n and a number o f theorem s d is c u s s e d r e c e n tly i n s t a t i s t i c a l l i t e r a t u r e a r e c o r o l l a r i e s o f one o f our th e o re m s. I n th e second p a r t o f t h i s p ap er th e th e o ry o f t e s t i n g s t a t i s ­ t i c a l h y p o th e se s w here th e form o f th e d i s t r i b u t i o n fu n c tio n s i s t o t a l l y unknown and o n ly c o n tin u ity i s assum ed, w i l l be d is c u s s e d .

The e x a c t

e x te n s io n o f th e li k e lih o o d r a t i o c r i t e r i o n to t h i s case w i l l be g iv e n . A pproxim atio n s to th e a p p l ic a t io n o f t h i s c r i t e r i o n in two problem s w ill be p ro p o se d , one o f w hich a p p lie s th e r e s u l t s m entioned above. L a s t ly , ,I i n c o n n e c tio n w ith th e second problem , a c o m jin a tttp a l problem w i l l be s o lv e d w hich i s new and has i n t e r e s t p er s e . 2. P a r t i t i o n s o f a s in g le i n t e g e r . L e t n be a p o s i t i v e in te g e r and

■1

be any sequence of p o s i t i v e in te g e r s

< t-

( I

-

, ............... , * )

1, 1

i/ok.».ine S

L -i

and s may be any in te g e r from I

VI

. Two sequences A w hich have

d i f f e r e n t elem en ts o r th e same elem en ts a rra n g e d i n d i f f e r e n t o rd e r a re H —I sequences A.

Z

We s h a l l co n sid e r th e sequence A as a s to c h a s t ic v a r ia b le

and a s s ig n t o a l l sequences A th e same p r o b a b i l i t y , w hich i s th e r e f o r e ^ -* 4 1 L et Yj be th e number o f elem en ts CU in A which eq u al

j

so

i s a s to c h a s t ic v a r ia b le . L et

— y\

be an in te g e r

j o i n t d i s t r i b u t i o n o f th e s to c h a s t ic v a r ia b le s i s g iven as fo llo w s :

} ^

The p r o b a b i l i t y t h a t

‘fclia‘fc "f* ^ .

Then th e

Y \ Jr.

^ ^ ( i s

I/ Z f ..

Kj

iS

(

M

' W

w here th e in n e r summation i s c a r r i e d out o v er a l l s e t s o f n o n -n e g a tiv e in te g e r s

V

(Z't'j i t +

V

"

+ • ■• +

S u c k ^

' t ~

l K+

>( K * . | ) 't'

f

' ■' ' +

r »u

^

T "

4 ( J .i)

(The

+

t-'

o f c o u rs e , a re n o n -n e g a tiv e i n t e g e r s . ) -

2-

' -

M

Let

1/u

v -Z .n «, = 1

and

(K -K l) ^ ' so t h a t O f

c= K + l

O p/ . . a r e b o th s to c h a s t ic v a r ia b le s . (£ + 0 p r o b a b i l i t y t h a t a t th e same tim e

V\ *4 i

(2.fJ and ^

and

r )

V

i j

a

is g iv e n by f ^ . l )

M

The

G ' 1, ■■■>K) Z k +1) •with th e r e s t r i e t i o n

V-)+ " +^ * ^ K+V

( a f t e r rem oval o f p a r e n th e s e s ) i s . asy m p to -

vm tim e s 4*V th\e l1 ao os4t- +term o Pf

4" 4 «n ±l\

H ence, f o r a l l

( > 2 + */

**

A ^ 2-

"K^ s u f f i c i e n t l y l a r g e ,

(l«J

i

a

33-

)

^

and

m u l tip l ie s th e f i r s t by

(? -« J V and th e t h i r d term o f th e r i g h t member by

(a-W

^

J/ - -

’ )

A

A '

I t i s easy t o see t h a t f d r la r g e b u t f ix e d

and a l l "Vt g r e a te r th a n

a low er bound w hich i s a f u n c tio n o f f o n l y , th e e x p re s s io n ^ i s l e s s th a n th e e x p r e s s io n

»

- 2. "3

Hence, in view o f ^ 3 ' ^-'6

th e sum o f th e f i r s t and t h i r d term s o f th e r i g h t member o f (j^ * v

fo r

t —



i s n e g a tiv e .

th e second te rm o f th e r i g h t member o f from

to

I

*

Now c o n s id e r what happens to

0J.r)

U

when

goes

I t i s m u ltip lie d by

O a a ) ( > - / “• - 1 1 w hich, a ls o f o r la r g e b u t fix e d and a l l bound w hich i s a f u n c tio n o f J&.

»

l a r g e r th a n a low er

o n ly , i s e a s i l y seen t o be l e s s th a n

C o nsequently '

( j . n j

Cs-rJ

I t can be se e n w ith o u t d i f f i c u l t y t h a t such a p assag e o f

to th e n e x t h ig h e r in d e x i s always accom panied by m u l t i p l i c a t i o n by e x p re s s io n s s im i la r to

, and

(

3

f o r w hich s im i la r i n e q u a l i t i e s h o ld and t h a t co n seq u en tly c

( j ztj

0 £

tr* (j~lc

and f o r s im i la r re a s o n s

-34«


0 / V

e tc .

-o

J

I t i s c l e a r t h a t th e le n g th s and lo c a tio n o f th e in t e r v a ls des­

c rib e d a r e im m a te ria l, p ro v id ed o n ly t h a t th e y do n o t o v e rla p . th e d i s t r i b u t i o n s o f

X T an d

A lso

w ith in each i n t e r v a l a r e im m a te ria l,

p ro v id e d only t h a t th e y a r e c o n tin u o u s .

A ll t h a t m a tte r s f o r f in d in g

i s t h a t th e number and th e o rd e r o f th e d is ­ ju n c t

i n t e r v a l s s h a l l be th e same as th o s e o f th e ru n s in

( i . e . , i n t e r v a l s o f p o s iti v e p r o b a b i l i t y fo r w ith i n t e r v a l s o f p o s i t i v e p r o b a b il it y f o r t e r v a l s o f p o s itiv e p r o b a b i l i t y f o r ^

^

and f o r

p e c t iv e l y th e number o f ru n s o f th e elem en t o f th e elem ent

0

\J

,

)^~

must a l t e r n a t e

*

The number o f in ­

T

m ust equal r e s ­

&ud th e number o f nuns

| , and th e p r o b a b il it y o f th e f i r s t i n t e r v a l on th e

l e f t s h a l l be p o s i t i v e f o r ^

o r fo r T

-4 2 -

acc o rd in g a s th e f i r s t run

in

V

V

i s o f elem en ts

0

o r o f elem en ts

| , w ith th e same r e ­

l a t i o n o b ta in in g betw een th e l a s t i n t e r v a l on th e r i g h t and th e l a s t ru n i n

\/

) and th e p r o b a b i l i t y o f th e s e i n t e r v a l s .

L et

^

be th e so u g h t f o r p r o b a b i l i t y o f th e i n t e r v a l w hich co rresp o n d s to th e

* Tfc I —*

ru n o f elem en ts

0

and

O .

th e p r o b a b il it y o f th e i n -

t e r v a l w hich co rresp o n d s t o th e I — ru n o f elem en ts o rd e r t o o b ta in

\wl

{

.

In

, i t i s n e c e s s a ry t h a t th e elem en ts c o n s ti tu ti n g

each ru n s h a l l f a l l i n t o i t s c o rresp o n d in g i n t e r v a l .

Then c l e a r l y

by t h e m u ltin o m ia l theorem

C H P { V ; ( I * , A,

*

I t-

L w here

(/ 5

2.

\

and w h ere, when

i s f ix e d , th e p ro d u ct

%

w ith r e s p e c t to m e n t. to th e

J

i s ta k e n ov er a l l runs o f th e c o rresp o n d in g e l e ­

The r i g h t member o f v • I C

i s to be maximized w ith r e s p e c t

, s u b je c t o f co u rse to th e c o n s tr a i n ts

( p i Then i t may e a s i l y be - r e r if i e d t h a t th e maximum o ccu rs when

GrtJ

pr =

( J * /, p)

'^/i

J

F o r , a f t e r m u ltip ly in g by a c o n s ta n t and ta k in g th e lo g a rith m we i n ­ tro d u c e two Lagrange m u l t i p l i e r s

f/L /

43

'

and

JU, /

so t h a t th e max-

im iz in g

'p . • a r e g iv e n b y th e e q u a tio n s [ JT, 3 J and I u\ V J th o s e o b ta in e d by e q u a tin g to zero a l l th e p a r t i a l d e r iv a tiv e s o f

j

I 1

- 'J

The l a t t e r a r e th e r e f o r e

;

a fo r a l l

J

, whence

O ' . - J

(T4J

' fo llo w s .

I t i s e a sy to se e th a t t h e

extremum th u s o b ta in e d i s a maximum and a ls o an a b s o lu te maximum. s o u g h t-f o r s t a t i s t i c

-r(Jl

a f t e r th e r e s u l t s

i s th e n th e r i g h t member o f

have been i n s e r t e d .

I t may be s im p lif ie d

by rem oving a l l f a c t o r s w hich a r e fu n c tio n s o n ly o f 'Vb ( s in c e th e s e w i l l th e n be th e same f o r a l l

Cr , r )

4

The

1/ y )

and

~l/b.

r e c a l l i n g th a t

* l/

°

I t w i l l be co n v e n ie n t to ta k e th e lo g a rith m o f th e r e s u l t i n g e x p r e s s io n , so t h a t w ith \

, we had re v e rse d

• we woui^ have o b tain ed th e p e r-

th e r o le s o f th e

X 11 *

m u tation

. I t i s easy to see th a t any s t a t i s t i c , say

r> I

snd

^

J

J

* ff

used to t e s t th e n u ll h y p o th e sis, must be a fu n c tio n only o f

iro j=

w ith th e added p ro v iso t h a t

, „

(F ) Under th e fI - "Vj ^

(The rank c o r r e la tio n c o e f f ic ie n t is such a s t a t i s t i c ) . o

(

^

^ th e n

I

Q i+ i '+ i j

;

Cl t J if

i

i

>i

The ru n w i l l be c a lle d an asc e n d in g ru n or a descen d in g ru n a c c o rd ­ in g as

+

A ru n o f le n g th

|

^

^

i s o f e i t h e r ty p e , a t p le a s u r e .

£=

'/ ^ (e

The f i r s t ru n i s

, th e l a s t ^

2.

* > ry ^ is

3 ^2 - a d e scen d in g r u n o f le n g th

( a ru n o f le n g th o n e, a. , * Q > J/ , . ^ r h r . b tL T iOv^— ^ M y ^ J i s a de gene r a t ^ f u n c t i on such t h a t th e

r e l a t i o n betw een

X

i s f u n c tio n a l ( t h i s i s a s p e c ia l

case o f s to c h a s t ic r e l a t i o n s h i p ) » That i s where

F o r exam ple, l e t

% 3/

second ^

a n ascen d in g ru n o f le n g th two j ^ t h r e e , and

~~ 1 ___________ o * 4 - | .

t o say , X

*

f> ( X

f\ (— \ v ) which c o n ta in e x a c tly ^ L

ru n s. C o n sid e r, f o r 2 3 4 6 5 1 .

exam ple, fo r th e case ^

^

, th e sequence

Vie s h a ll say t h a t t h i s sequence c o n ta in s th e "con­

ta c ts " (2 ,3 ) , ( 3 , 4 ) , ( 6 , 5 ).

In g e n e r a l, a c o n ta c t i s d e fin e d as

th e ju x t a p o s i t i o n , in t h e sequence

o f c o n s e c u tiv e num bers, w hether I f i i . i s t h e number o f ru n s and

in a scen d in g or d escen d in g o r d e r .

th e number o f c o n ta c ts i n a sequence

th e n o b v io u sly

(1,|J L et ^ q be th e sequence

................. ''Yv--' o f th e f i r s t y\ \ y- i n -

1, 2,

te g e r s i n aso en d in g o r d e r .

The

''Yv.— |

c o n ta c ts o f t h i s sequence * of

R

c o n t a c ts , t h u s : (1 , 2 ), ( 2 , 3 ) , ................................................ me a n in g '^*1

(" h — | ,

TV

- iw............................................upp........ifn

of th e c o n ta c ts which c o n s ti tu te th e sequence

JV

) .i Suppose a r e s e le c te d

The re m a in in g

t a c t s form the com plementary s e t

* A fte r t h i s s e le c tio n th e s e ­

quence

| —

con-

in some manner t o form th e s e t Q .

R * may be c o n s id e re d a sequence of th e type o f th e se­

quences ^

o f S e c tio n 5 w ith th e members of Q

o f th e elem ents th e elem en ts I

w rite i t as

0

and t h e members o f

. Tflhen

R*€>J

p la y in g th e r o le

p la y in g th e r o l e of

is c o n sid ere d i n t h i s manner we w i l l



The d e f in itio n o f a run o f Section. 6"

56'



imehts W ts

th e Member

ana

\/N h e s j

/M .s e 5r a te -

£ irfo o tio a E> as a p p lie d to seq u en ces c a b le to

i s now a p p li­

We w i l l c a l l any su ch ru n o f th e members o f

o r of

(3

a group. We w ish f i r s t t o answ er th e fo llo w in g q u e s tio n :

O

b e s e le c te d from among th e

t h a t i t w i l l c o n ta in ^

A

I f , f o r a g iv e n d iv id e d in | ^ 0 j

-N

=

t

elem en ts o f b * [0 1

members a rra n g e d i n

•/

so

in C groups?

7

A '

, C be th e number o f groups i n t o which U

R * C o J , i t is c le a r th a t or

In how many

.

C— L

is

can e q u a l o n ly

Hence o n ly f o u r s i t u a t i o n s can a r i s e , as f o l ­

lo w s: a)

t /

. o f elem ents o f 0

th e r e f o r e composed y

elem en ts can be d iv id e d i n t o t

i s th e c o e f f i c i e n t o f

The f i r s t group in .

The number of ways i n w hich ru n s o f th e ty p e o f S e c tio n 2

in th e p u re ly fo rm al e x p an sio n o f

+ - -■ j and i s th e r e f o r e

rz-i \I \ .

-

(y -j

i

V t-I I elem en ts can be d iv id e d i n t o w ays.

R * ( 0 ) is

C

j?

. S im ila r ly 'V t *—I v, ■ f'+ .'l - ' O "f"' I ru n s i n I £

Hence t h i s s i t u a t i o n w i l l a r i s e in

jJ

^~J

ways.. * /

b)

L

occur m



C—I



By a s im ila r argum ent a s ab o v e, t h i s can

L_ l -

-67-

=;

c*

^

ti* jv**-

•e j&.

k

^ r o

. T k r w*

0

_y/which con­ ta in

e x a c tly ^

c o n t a c ts .

As was s a id b e f o r e , th e t o t a l of th e num­

b e r of sequences in ea c h is



-5 9 -

L et

"^e s e ^

^e

sequences i n a l l th e s e f a m i l i e s , w ith eac h sequence i n

o u n t-

ed as many tim es a s th e number o f f a m ilie s i n w hich i t o c c u rs . e r y sequence i n

has th e

g e n e ra te d i t , b u t a f t e r p erm u tin g Hence e v e ry sequence in by C M

m

( 3 w hich

o th e r c o n ta c ts may s t i l l e x i s t .

h a s a t l e a s t ^ c o n ta c ts and th e r e f o r e

, a t m ost

has e x a c tly

c o n ta c ts o f th e s e t

ru n s .

C le a r ly , a sequence w hich

c o n ta c ts o c c u rs e x a c tly once i n

, s in c e i t

can a p p e a r o n ly i k th e fa m ily g e n e ra te d by th e s e t t a c t s and i n no o th e r fa m ily .

Q

of its

tim e s in

f o r i t w i l l ap p ea r once i n each fa m ily g e n e ra te d by a s e t c o n s is ts o f one o f t h e ^ ^ M

s e le c tio n s o f

(^, ■+■2^ J

fitAtim e s ,

and so f o r t h .

Q

which

c o n ta c ts from among

c o n ta c ts , and in no o th e r fa m ily .

quence w hich has e x a c tly

con­

A sequence w hich has e x a c tly

c o n ta c ts w i l l ap p ea r e x a c tly

its ( e

Ev­

S im ila r ly each s e ­

c o n ta c ts w i l l ap p ear i n

«(«/

We t h e r e f o r e o b ta in , i n view o f C M ,

i

C 2/ The system o f

l i n e a r e q u a tio n

d eterm in es th e q u a n titie s

N o | ,