Let $(\Omega, \mathscr{F}, \mathbb{P})$ be a probability space. There are two different definitions of the independence of random variables $X_1, X_2, ...$ on $(\Omega, \mathscr{F}, \mathbb{P})$:

  1. The $\sigma$-algebras $\sigma(X_1), \sigma(X_2), ...$ are independent, i.e. for distinct indices $k_1, k_2, ..., k_m$, $\sigma(X_{k_1}), \sigma(X_{k_2}), ..., \sigma(X_{k_m})$ are independent.

  2. Given Borel sets $B_1, B_2, ... $ and distinct indices $j_1, j_2, ..., j_p$,

$$\prod_{j=j_1}^{j_p} P(X_j \in B_j) = P(\bigcap_{j=j_1}^{j_p} (X_j \in B_j))$$

Why are these two definitions equivalent?

What I tried:

For distinct indices $k_1, k_2, ..., k_m$,

$\sigma(X_{k_1}), \sigma(X_{k_2}), ..., \sigma(X_{k_m})$ are independent.

$$\iff P(\bigcap_{k=k_1}^{k_m} A_k) = \prod_{k=k_1}^{k_m} P(A_k)$$


$$A_k \in \sigma(X_k) = \{ X_k^{-1}(B_k) | B_k \in \mathscr{B} \}$$

It looks like

$$P(\bigcap_{k=k_1}^{k_m} A_k) = \prod_{k=k_1}^{k_m} P(A_k)$$

is the same as

$$\prod_{j=j_1}^{j_p} P(X_j \in B_j) = P(\bigcap_{j=j_1}^{j_p} (X_j \in B_j))$$

Does that end the proof? Doesn't seem very rigorous. Can I somehow let or conclude that $m = p$ and $(k_1, ..., k_m) = (j_1, ..., j_p)$?


From definition 1, we can fix $m \in \mathbb{N}$, $k_1, k_2, \dots, k_m \in \mathbb{N}$ and $B_{k_1}, B_{k_2}, \dots, B_{k_m} \in \mathscr{B}$ s.t. $$P(\bigcap_{k=k_1}^{k_m} A_k) = \prod_{k=k_1}^{k_m} P(A_k)$$

Since $$\prod_{j=j_1}^{j_p} P(X_j \in B_j) = P(\bigcap_{j=j_1}^{j_p} (X_j \in B_j))$$ holds for all $p \in \mathbb{N}$, all distinct indices $j_1, \dots, j_p \in \mathbb{N} $ and all $B_{j_1}, B_{j_2}, \dots, B_{j_p} \in \mathscr{B}$ and since $m \in \mathbb{N}$, $k_1, k_2, \dots, k_m \in \mathbb{N}$ and $B_{k_1}, B_{k_2}, \dots, B_{k_m} \in \mathscr{B}$ we can choose $p = m$, $$j_i = k_i$$ and $B_{j_i} = B_{k_i}$ for $i = 1, \dots p \ ( = m )$. QED


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.