26.6. REDUCTION TO FINITE DIMENSIONS 727

26.6 Reduction to Finite DimensionsLet E be a Banach space and let g ∈ (E ′)n . Then for x ∈ E, g ◦ x is the vector in Fn whichequals (g1 (x) ,g2 (x) , · · · ,gn (x)).

Theorem 26.6.1 Let Xi be a random variable having values in E a real separableBanach space. The random variables {Xi}i∈I are independent if whenever

{i1, · · · , in} ⊆ I,

mi1 , · · · ,min are positive integers, and gmi1, · · · ,gmin

are respectively in

(M)mi1 , · · · ,(M)min

for M a dense subspace of E ′,{gmi j◦Xi j

}n

j=1are independent random vectors having

values in Rmi1 , · · · ,Rmin respectively.

Proof: It is necessary to show that the events X−1i j

(Bi j

)are independent events when-

ever Bi j are Borel sets. By Lemma 26.5.1 and the above Lemma 26.5.2, it suffices to verifythat the events

X−1i j

(g−1

mi j

(Cα⃗,mi j

))=(gmi j◦Xi j

)−1(Cα⃗,mi j

)are independent where Cα⃗,mi j

are the cones described in Lemma 26.5.2. Thus

α⃗=(αk1 , · · · ,αkm

), Cα⃗,mi j

=

mi j

∏i=1

(−∞,αki ]

But this condition is implied when the finite dimensional valued random vectors gmi j◦Xi j

are independent. ■The above assertion also goes the other way as you may want to show.

26.7 0,1 LawsI am following [55] for the proof of many of the following theorems. Recall the set of ω

which are in infinitely many of the sets {An} is ∩∞n=1∪∞

m=n Am. This is in ∩∞n=1∪∞

m=n Am ifand only if for every n there exists m≥ n such that it is in Am.

Theorem 26.7.1 Suppose An ∈ Fn where the σ algebras {Fn}∞

n=1 are indepen-dent. Suppose also that ∑

∞k=1 P(Ak) = ∞. Then P(∩∞

n=1∪∞m=n Am) = 1.

Proof: It suffices to verify that P(∪∞

n=1∩∞m=n AC

m)= 0 which can be accomplished by

showing that P(∩∞

m=nACm)= 0 for each n. The sets

{AC

k

}satisfy AC

k ∈ Fk. Therefore,noting that e−x ≥ 1− x,

P(∩∞

m=nACm)

= limN→∞

P(∩N

m=nACm)= lim

N→∞

N

∏m=n

P(AC

m)

= limN→∞

N

∏m=n

(1−P(Am))≤ limN→∞

N

∏m=n

e−P(Am)

= limN→∞

exp

(−

N

∑m=n

P(Am)

)= 0. ■

26.6. REDUCTION TO FINITE DIMENSIONS 72726.6 Reduction to Finite DimensionsLet E be a Banach space and let g € (E’)". Then for x € E, gox is the vector in F” whichequals (g1 (x) , 82 (x) ,-++ ,8n(x)).Theorem 26.6.1 Lez X; be a random variable having values in E a real separableBanach space. The random variables {X;};<; are independent if whenever{iiss sin} CT,Mj, ,*** ,Mj;, are positive integers, and Inj, 97° + Im;, VE respectively inmy Mi,(My ++», (M)"™nfor M a dense subspace of E', { 9m, oXi,} are independent random vectors havingJvalues in R™ 1 ,--» , Rn respectively.Proof: It is necessary to show that the events Xj! (Bi,) are independent events when-ever Bi; are Borel sets. By Lemma 26.5.1 and the above Lemma 26.5.2, it suffices to verifythat the events 7"(1 (con) =(am 0%)" (Com)are independent where Cg, are the cones described in Lemma 26.5.2. ThusJ3ia= (Ox, 5-°° , Hkin ) ’ Cami, = (—0, OK; |i=lBut this condition is implied when the finite dimensional valued random vectors g,,,_° Xi;Jare independent. MfThe above assertion also goes the other way as you may want to show.26.7 0,1 LawsI am following [55] for the proof of many of the following theorems. Recall the set of @which are in infinitely many of the sets {A,} is N%_; U7, Am- This is in N_, Urn Am ifand only if for every n there exists m > n such that it is in A».Theorem 26.7.1 Suppose A, € Fn where the o algebras {Fn};,_1 are indepen-dent. Suppose also that Yy_, P (Ax) = °°. Then P(M?_; Un Am) = 1.Proof: It suffices to verify that P (Ue_ 1 On=n AC) = 0 which can be accomplished byshowing that P(N%;_,AG,) = 0 for each n. The sets {Af} satisfy AC € F,. Therefore,noting that e~* > 1 —x,NP(MjinAS) = fim P (OM ,AS) = Jim TTP (4S)m=nN N= lim J] (1—P(A,)) < lim TT 4“N-e0 m=n N+ =nN-veo m=nNlim exp (- y? Pldn)) =0. i