528 CHAPTER 20. PRODUCT MEASURES

thenν t1···tn (Ft1 ×·· ·×Ftn) = νs1···sp

(Gs1 ×·· ·×Gsp

)(20.1)

where if si = t j, then Gsi = Ft j and if si is not equal to any of the indices tk, then Gsi = Msi .Then for E defined in Notation 20.3.1, there exists a probability measure P and a σ algebraF = σ (E ) such that (∏t∈I Mt ,P,F ) is a probability space. Also there exist measurablefunctions, Xs : ∏t∈I Mt →Ms defined for s ∈ I as Xsx≡ xs such that for each (t1 · · · tn)⊆ I,

ν t1···tn (Ft1 ×·· ·×Ftn) = P([Xt1 ∈ Ft1 ]∩·· ·∩ [Xtn ∈ Ftn ])

= P

((Xt1 , · · · ,Xtn) ∈

n

∏j=1

Ft j

)= P

(∏t∈I

Ft

)(20.2)

where Ft = Mt for every t /∈ {t1 · · · tn} and Fti is a Borel set. Also if f is a nonnegative

function of finitely many variables, xt1 , · · · ,xtn , measurable with respect to B(

∏nj=1 Mt j

),

then f is also measurable with respect to F and∫Mt1×···×Mtn

f (xt1 , · · · ,xtn)dν t1···tn =∫

∏t∈I Mt

f (xt1 , · · · ,xtn)dP (20.3)

Proof: Let E be the algebra of sets defined in the above notation. I want to define ameasure on E . For F ∈ E , there exists J such that F is the finite disjoint union of setsof RJ . Define P0 (F ) ≡ νJ (πJ (F )) . Then P0 is well defined because of the consistencycondition on the measures νJ . P0 is clearly finitely additive because the νJ are measuresand one can pick J as large as desired to include all t where there may be something otherthan Mt . Also, from the definition,

P0 (Ω)≡ P0

(∏t∈I

Mt

)= ν t1 (Mt1) = 1.

Next I will show P0 is a finite measure on E . After this it is only a matter of using theCaratheodory extension theorem to get the existence of the desired probability measure P.

Claim: Suppose En is in E and suppose En ↓ /0. Then P0 (En) ↓ 0.

Proof of the claim: If not, there exists a sequence such that althoughEn ↓ /0,P0 (En) ↓

ε > 0. Let En ∈ EJn . Thus it is a finite disjoint union of sets of RJn . By regularity of themeasures νJ , which follows from Lemmas 9.8.4 and 9.8.5, there existsKJn ⊆En such that

νJn (πJn (KJn))+ε

2n+2 > νJn (πJn (En))

Thus P0 (KJn)+ε

2n+2 ≡ νJn (πJn (KJn))+ε

2n+2 > νJn (πJn (En))≡ P0 (E

n) .The interestingthing about theseKJn is: they have the finite intersection property. Here is why.

ε ≤ P0(∩m

k=1KJk

)+P0

(Em \∩m

k=1KJk

)≤ P0

(∩m

k=1KJk

)+P0

(∪m

k=1Ek \KJk

)< P0

(∩m

k=1KJk

)+

∑k=1

ε

2k+2 < P0(∩m

k=1KJk

)+ ε/2,

and so P0(∩m

k=1KJk

)> ε/2. In considering all theEn, there are countably many entries in

the product space which have something other than Mt in them. Say these are {t1, t2, · · ·} .

528 CHAPTER 20. PRODUCT MEASURESthenViperty (Fay X00 X Fi) = Vosy-5p (Gop X00 X Gs,) (20.1)where if s; = tj, then Gs, = Fi, and if s; is not equal to any of the indices t,, then Gs, = Ms,.Then for & defined in Notation 20.3.1, there exists a probability measure P and a o algebraF =0(&) such that (Yre;M,P, F) is a probability space. Also there exist measurablefunctions, Xs : Trey Mr — Ms defined for s € I as X;x =x, such that for each (t+++th) CT,Vi --ty (Fi, x ++ X Fi, ) = P(X, € F,|-+- 0 |X, € F,,])= (ci ,X,,) € 11") -o(T1s) (20.2)j=ltelwhere F, = M, for every t ¢ {t,---t,} and F,, is a Borel set. Also if f is a nonnegativefunction of finitely many variables, x;,,-++ ,Xt,, measurable with respect to B (1-1 M, i)then f is also measurable with respect to F# and| Fins tg) Dim = [oF Oye vt dP (20.3)My, XM iq TresProof: Let & be the algebra of sets defined in the above notation. I want to define ameasure on &. For F' € &, there exists J such that F' is the finite disjoint union of setsof Z;. Define Py (F') = v;(a;(F)). Then Pp is well defined because of the consistencycondition on the measures v;. Po is clearly finitely additive because the v; are measuresand one can pick J as large as desired to include all t where there may be something otherthan M,. Also, from the definition,Po (Q) = Po (1) =v, (M,) =1.telNext I will show Pp is a finite measure on &. After this it is only a matter of using theCaratheodory extension theorem to get the existence of the desired probability measure P.Claim: Suppose E” is in & and suppose E” | 0. Then Po (E”) | 0.Proof of the claim: If not, there exists a sequence such that although EB” | 0, Po (E”) |€ > 0. Let BE” € &,. Thus it is a finite disjoint union of sets of Z,. By regularity of themeasures V;, which follows from Lemmas 9.8.4 and 9.8.5, there exists ky, C E” such thatE nNVin (Ty (Bn) + Sapa > Vn (Tn E"))Thus Po (KC y,) + 5257 = Vay, (Ty (I I,)) + saz > Vs, (1s, (E")) = Po (E") The interestingthing about these K’;, is: they have the finite intersection property. Here is why.€ < P(N, Ky,) +P) (B"\ 0, Ky) <P (ON, Ky) +P (Ui B*\ Ky)< PYM) + YD sea <Po(ML Ky) + €/2.k=1and so Py (AL, K a) > €/2. In considering all the E”, there are countably many entries inthe product space which have something other than M, in them. Say these are {f,f,---}.