Independent $\sigma$-algebras using $\pi$-$\lambda$-theorem

Question

Let $\mathcal{E}_1, ...,\mathcal{E}_n$ be collections of measurable sets on $(\Omega,\mathcal{F},P)$, each closed under intersection. Suppose \begin{align*} P(A_1\cap...\cap\ A_n)=P(A_1)\cdot ... \cdot P(A_n), \end{align*} for all $A_i \in \mathcal{E}_i$ for $1 \leq i \leq n$.

Now I want to show that the $\sigma$-algebras $\sigma(\mathcal{E}_i)$ for $1 \leq i \leq n$ are independent, using an application of the $\pi$-$\lambda$-theorem.

Since $\mathcal{E}_i$ for $1 \leq i \leq n$ are closed under intersection, each $\mathcal{E}_i$ is a $\pi$-system. Now, for me it is unclear how to define a $\lambda$-system and how to apply the $\pi$-$\lambda$-theorem.

Answer 1

Fix $A_i,\forall i=2,\cdots,n$, define $G=\{B\in \sigma(\mathcal{E}_1)|P(B\cap\cdots\cap\ A_n)=P(B)\cdot \cdots \cdot P(A_n)\}$,

By assumption $\mathcal{E}_1\subset G$, it suffices to show $G$ is a $\lambda$-system by definition.

Then by $\pi-\lambda$ theorem, we have $\sigma(\mathcal{E}_1)\subset G$, this shows given $\mathcal{E}_1, \cdots,\mathcal{E}_n$ independent, we have $\sigma(\mathcal{E}_1), \cdots,\mathcal{E}_n$ independent, hence $\mathcal{E}_2, \cdots,\mathcal{E}_n,\sigma(\mathcal{E}_1)$ independent. THen repeart

Answer 2

Since $\mathcal E_i$ are $π$-systems and independent, then you know (if, no there is a sketch of the proof in the end of this answer) that also the induced Dynkin systems $δ(\mathcal E_i)$ are independent. No, since the $\mathcal E_i$'s are $π$-systems, Dynkin's theorem states that $$δ(\mathcal E_i)=σ(\mathcal E_i)$$ which completes the proof.

---

Statement: If $\mathcal E_i,\ i \in I$ are independent in $(Ω,\mathcal F, P)$, then the induced Dynkin systems $δ(\mathcal E_i), i\in I$ are also independent.

Sketch of the proof: Define $$\mathsf E_1=\{A\in \mathcal F:P(AA_2\dots A_n)\}=P(A)P(A_2)\dots P(A_n), \forall A_i\in\mathcal E_i \cup Ω, i=2,\dots, n\}$$ i.e. $\mathsf E_1$ is the set of all the "good sets" that are independent of $\mathcal E_2, \dots, \mathcal E_n$ (so $\mathsf E_1$ is the maximal class such that $\mathsf E_1$ and $\mathcal E_2, \dots, \mathcal E_n$ are independent. No, show that $\mathsf E_1$ is Dynkin system (about two pages of calculations in my textbook...) which by definition contains $\mathcal E_1$. Hence the minimality of the induced Dynkin system $δ(\mathcal E_1)$ yields $δ(\mathcal E_1)\subseteq \mathsf E_1$ which completes the proof.