26.4. INDEPENDENT EVENTS AND σ ALGEBRAS 721

are equal because you have no more than l sets different than Aik . The only remaining caseis where some Bik = AC

ik. Say Bim+1 = AC

im+1for simplicity.

P(∩m+1

k=1 Bik

)= P

(AC

im+1∩∩m

k=1Bik

)= P

(∩m

k=1Bik

)−P

(Aim+1 ∩∩

mk=1Bik

)Then by induction,

=m

∏k=1

P(Bik

)−P

(Aim+1

) m

∏k=1

P(Bik

)=

m

∏k=1

P(Bik

)(1−P

(Aim+1

))= P

(AC

im+1

) m

∏k=1

P(Bik

)=

m+1

∏k=1

P(Bik

)thus proving it for l +1. ■

This motivates a more general notion of independence in terms of σ algebras.

Definition 26.4.3 If {Fi}i∈I is any set of σ algebras contained in F , they are saidto be independent if whenever Aik ∈Fik for k = 1,2, · · · ,m, then

P(∩m

k=1Aik

)=

m

∏k=1

P(Aik

).

A set of random variables {X i}i∈I is independent if the σ algebras {σ (X i)}i∈I are in-dependent σ algebras. Here σ (X) denotes the smallest σ algebra such that X is mea-surable. Thus σ (X) =

{X−1 (U) : U is a Borel set

}. More generally, σ (X i : i ∈ I) is the

smallest σ algebra such that eachX i is measurable.

Note that by Lemma 26.4.2 you can consider independent events in terms of indepen-dent σ algebras. That is, a set of independent events can always be considered as eventstaken from a set of independent σ algebras. This is a more general notion because here theσ algebras might have infinitely many sets in them.

Lemma 26.4.4 Suppose the set of random variables, {X i}i∈I is independent. Alsosuppose I1 ⊆ I and j /∈ I1. Then the σ algebras σ (X i : i ∈ I1) , σ (X j) are independent σ

algebras.

Proof: Let B ∈ σ (X j) . I want to show that for any A ∈ σ (X i : i ∈ I1) , it follows thatP(A∩B) = P(A)P(B) . Let K consist of finite intersections of sets of the formX−1

k (Bk)where Bk is a Borel set and k ∈ I1. Thus K is a π system and σ (K ) = σ (X i : i ∈ I1) .This is because it follows from the definition that σ (K )⊇ σ (X i : i ∈ I1) because σ (K )contains all X−1

i (B) for B Borel. For the other inclusion, the right side consists of all setsX−1

i (B) where B is a Borel set and so the right side, being a σ algebra also must containall finite intersections of these sets which means σ (X i : i ∈ I1) must contain K and soσ (X i : i ∈ I1)⊇ σ (K ).

Now if you have one of these sets of the form A = ∩mk=1X

−1k (Bk) where without

loss of generality, it can be assumed the k are distinct since X−1k (Bk) ∩X−1

k

(B′k)=

X−1k

(Bk ∩B′k

), then

P(A∩B) = P(∩m

k=1X−1k (Bk)∩B

)= P(B)

m

∏k=1

P(X−1

k (Bk))

= P(B)P(∩m

k=1X−1k (Bk)

).

26.4. INDEPENDENT EVENTS AND o ALGEBRAS 721are equal because you have no more than / sets different than A;,. The only remaining caseis where some B;, = Af. Say B;,,,, =A‘, for simplicity.AL Bi)im+1P (1 Bix) =P (46, Bi) i P(N Bix) —P(AThen by induction,T1? (Bi) Pigs) FL? Bi) = FL (i) (1 P Aig)m+1(a%.,) TP.) = TPG)thus proving it for/+1.This motivates a more general notion of independence in terms of o algebras.Definition 26.4.3 If { Fi} je, is any set of 6 algebras contained in F, they are saidto be independent if whenever Aj, € Fi, for k = 1,2,--- ,m, thenmP (MLA; ) — []? (i) :k=lA set of random variables {Xj},-, is independent if the o algebras {6 (Xj) }je, are in-dependent o algebras. Here o(X) denotes the smallest o algebra such that X is mea-surable. Thus 0 (X) = {x7! (U) :U is a Borel set} . More generally, o (Xj :i€ 1) is thesmallest 0 algebra such that each X ; is measurable.Note that by Lemma 26.4.2 you can consider independent events in terms of indepen-dent o algebras. That is, a set of independent events can always be considered as eventstaken from a set of independent o algebras. This is a more general notion because here theo algebras might have infinitely many sets in them.Lemma 26.4.4 Suppose the set of random variables, {X iticy is independent. Alsosuppose I, CI and j ¢ 1. Then the o algebras o (Xj; :i € 1), 6 (X ;) are independent oalgebras.Proof: Let B € o (X ;). I want to show that for any A € 0 (X;:1€1;), it follows thatP(ANB) =P(A)P(B). Let .% consist of finite intersections of sets of the form X i (Bx)where B; is a Borel set and k € l. Thus .% is a m system and 0 (.%) =o (X;:iEN).This is because it follows from the definition that o (4%) D o (X;:i€ I) because o (.%)contains all X 7! (B) for B Borel. For the other inclusion, the right side consists of all setsx>! (B) where B is a Borel set and so the right side, being a o algebra also must containUall finite intersections of these sets which means o (X;:i€J,) must contain % and so0o(X;:iE€h)DIo(#).Now if you have one of these sets of the form A = ML) X i (By) where withoutloss of generality, it can be assumed the k are distinct since X;,! (By) 1X;,' (Bi) =X,! (BOB), thenP(ANB) = P(NL,X;,' (Be) NB) =P(B) Tee (By)k=1= P(B)P(L) X;' (Bx))-