Dynkin system

A Dynkin system, named after Eugene Dynkin, is a collection of subsets of another universal set Ω satisfying a set of axioms weaker than those of 𝜎-algebra. Dynkin systems are sometimes referred to as 𝜆-systems (Dynkin himself used this term) or d-system. These set families have applications in measure theory and probability.

A major application of 𝜆-systems is the π-𝜆 theorem, see below.

Definition

Let Ω be a set, and let D be a collection of subsets of Ω (that is, D is a subset of the power set of Ω). Then D is a Dynkin system if

  1. Ω∈D;
  2. D is closed under complements of subsets in supersets: if A,B∈D and A⊆B, then B∖A∈D;
  3. D is closed under countable increasing unions: if A1⊆A2⊆A3⊆⋯ is an increasing sequence of sets in D then ⋃n=1∞An∈D.

It is easy to check that any Dynkin system D satisfies:

  1. ∅∈D;
  2. D is closed under complements in Ω: if A∈D, then Ω∖A∈D;
    • Taking A:=Ω shows that ∅∈D.
  3. D is closed under countable unions of pairwise disjoint sets: if A1,A2,A3,… is a sequence of pairwise disjoint sets in D (meaning that Ai∩Aj=∅ for all i≠j) then ⋃n=1∞An∈D.
    • To be clear, this property also holds for finite sequences A1,…,An of pairwise disjoint sets (by letting Ai:=∅ for all i>n).

Conversely, it is easy to check that a family of sets that satisfy conditions 4-6 is a Dynkin class. For this reason, a small group of authors have adopted conditions 4-6 to define a Dynkin system.

An important fact is that any Dynkin system that is also a π-system (that is, closed under finite intersections) is a 𝜎-algebra. This can be verified by noting that conditions 2 and 3 together with closure under finite intersections imply closure under finite unions, which in turn implies closure under countable unions.

Given any collection 𝒥 of subsets of Ω, there exists a unique Dynkin system denoted D{𝒥} which is minimal with respect to containing 𝒥. That is, if D~ is any Dynkin system containing 𝒥, then D{𝒥}⊆D~. D{𝒥} is called the Dynkin system generated by 𝒥. For instance, D{∅}={∅,Ω}. For another example, let Ω={1,2,3,4} and 𝒥={1}; then D{𝒥}={∅,{1},{2,3,4},Ω}.

Sierpiński–Dynkin's π-λ theorem

Sierpiński-Dynkin's π-𝜆 theorem: If P is a π-system and D is a Dynkin system with P⊆D, then σ{P}⊆D.

In other words, the 𝜎-algebra generated by P is contained in D. Thus a Dynkin system contains a π-system if and only if it contains the 𝜎-algebra generated by that π-system.

One application of Sierpiński-Dynkin's π-𝜆 theorem is the uniqueness of a measure that evaluates the length of an interval (known as the Lebesgue measure):

Let (Ω,ℬ,ℓ) be the unit interval [0,1] with the Lebesgue measure on Borel sets. Let m be another measure on Ω satisfying m[(a,b)]=b−a, and let D be the family of sets S such that m[S]=ℓ[S]. Let I:={(a,b),[a,b),(a,b],[a,b]:0<a≤b<1}, and observe that I is closed under finite intersections, that I⊆D, and that ℬ is the 𝜎-algebra generated by I. It may be shown that D satisfies the above conditions for a Dynkin-system. From Sierpiński-Dynkin's π-𝜆 Theorem it follows that D in fact includes all of ℬ, which is equivalent to showing that the Lebesgue measure is unique on ℬ.

Application to probability distributions

The π-𝜆 theorem motivates the common definition of the probability distribution of a random variable X:(Ω,ℱ,P⁡)→ℝ in terms of its cumulative distribution function. Recall that the cumulative distribution of a random variable is defined as FX(a)=P⁡[X≤a],a∈ℝ, whereas the seemingly more general law of the variable is the probability measure ℒX(B)=P⁡[X−1(B)] for all B∈ℬ(ℝ), where ℬ(ℝ) is the Borel 𝜎-algebra. The random variables X:(Ω,ℱ,P⁡)→ℝ and Y:(Ω~,ℱ~,P~)→ℝ (on two possibly different probability spaces) are equal in distribution (or law), denoted by X=𝒟Y, if they have the same cumulative distribution functions; that is, if FX=FY. The motivation for the definition stems from the observation that if FX=FY, then that is exactly to say that ℒX and ℒY agree on the π-system {(−∞,a]:a∈ℝ} which generates ℬ(ℝ), and so by the example above: ℒX=ℒY.

A similar result holds for the joint distribution of a random vector. For example, suppose X and Y are two random variables defined on the same probability space (Ω,ℱ,P⁡), with respectively generated π-systems ℐX and ℐY. The joint cumulative distribution function of (X,Y) is FX,Y(a,b)=P⁡[X≤a,Y≤b]=P⁡[X−1((−∞,a])∩Y−1((−∞,b])], for all a,b∈ℝ.

However, A=X−1((−∞,a])∈ℐX and B=Y−1((−∞,b])∈ℐY. Because ℐX,Y={A∩B:A∈ℐX, and B∈ℐY} is a π-system generated by the random pair (X,Y), the π-𝜆 theorem is used to show that the joint cumulative distribution function suffices to determine the joint law of (X,Y). In other words, (X,Y) and (W,Z) have the same distribution if and only if they have the same joint cumulative distribution function.

In the theory of stochastic processes, two processes (Xt)t∈T,(Yt)t∈T are known to be equal in distribution if and only if they agree on all finite-dimensional distributions; that is, for all t1,…,tn∈T,n∈ℕ, (Xt1,…,Xtn)=𝒟(Yt1,…,Ytn).

The proof of this is another application of the π-𝜆 theorem.

See also