Probabilistic proof for $\prod_{n} (1-p_n) > 0$ (Considering the $p_n$'s as parameters of random variables and using Borel-Cantelli and continuity)

385 Views Asked by At

Consider a sequence $\{p_n\}_{n \in \mathbb{N}}$ s.t. $p_n \in [0,1)$ and $\sum_{n=1}^{\infty} p_n < \infty$.

Prove $\prod_{n=1}^{\infty} (1-p_n) > 0$.


I think there's a way to do this without probability eg this or this. I mean there's no probability here originally. But I would like to try a probabilistic proof. Kinda like this or this


What I tried (I'm not sure if my lim, inf and sup statements are right):

Let $(\Omega, \mathscr{F}, \mathbb{P})$ be a probability space, and let $X_n$ be iid Bernoulli random variables with parameter $p_n$ and in $(\Omega, \mathscr{F}, \mathbb{P})$. Then,

$$\sum_{n=1}^\infty p_n<\infty \iff \sum_{n=1}^\infty P(X_n = 1) <\infty$$

By Borel-Cantelli Lemma 1 we have, $$P(\limsup(X_n = 1))=0$$

which is equivalent to $$P(\liminf(X_n = 0)) = 1$$

$$ \ $$

$$\to 1 =P(\bigcup_{N \ge 1} \bigcap_{n \ge N} (X_n=0))$$

By continuity of measure $\color{red}{\text{(I think?)}}$

$$= \lim_{N \to \infty}P(\bigcap_{n \ge N} (X_n=0))$$

By monotone convergence theorem $\color{red}{\text{(I think?)}}$

$$= \sup_{N \ge 1}P(\bigcap_{n \ge N} (X_n=0))$$

By independence,

$$ = \sup_{N \ge 1} \prod_{n=N}^{\infty} P(X_n=0)$$

$$= \sup_{N \ge 1} \prod_{n=N}^{\infty} (1-p_n)$$

$$\to \sup_{N \ge 1} \prod_{n=N}^{\infty} (1-p_n) = 1$$

$$\to \forall \epsilon > 0, \exists m \ge 1 \text{s.t.} \prod_{n=m}^{\infty} (1-p_n) \in (1-\epsilon, 1] \ \color{red}{\text{(I think?)}}$$

Since $p_n < 1 \ \forall n \in \mathbb{N}$,

$$\forall \epsilon \in (0,1), \exists m \ge 1 \text{s.t.} \prod_{n=m}^{\infty} (1-p_n) \in (1-\epsilon, 1] \ \color{red}{\text{(I think?)}}$$

Hence $\forall \epsilon \in (0,1), \exists m \ge 1 \text{s.t.}$

$$\prod_{n=1}^\infty(1-p_n)=\prod_{n=1}^{m-1}(1-p_n)\prod_{n=m}^{\infty}(1-p_n) > 0 \ \because$$

  1. $\prod_{n=1}^{m-1}(1-p_n) > 0 \because p_n < 1 \ \forall n \ge 1$

  2. $\prod_{n=m}^{\infty}(1-p_n) > 0 \because \prod_{n=m}^{\infty} (1-p_n) \in (1-\epsilon, 1]$ QED

Any mistakes? Again, I'm not sure if my lim, inf and sup statements are right.

1

There are 1 best solutions below

0
On BEST ANSWER

Let $(\Omega, \mathscr{F}, \mathbb{P})$ be a probability space, and let $X_n$ be iid Bernoulli random variables with parameter $p_n$ and in $(\Omega, \mathscr{F}, \mathbb{P})$. Then,

$$\sum_{n=1}^\infty p_n<\infty \iff \sum_{n=1}^\infty P(X_n = 1) <\infty$$

By Borel-Cantelli Lemma 1 we have, $$P(\limsup(X_n = 1))=0$$

which is equivalent to $$P(\liminf(X_n = 0)) = 1$$

$$ \ $$

$$\to 1 =P(\bigcup_{N \ge 1} \bigcap_{n \ge N} (X_n=0))$$

By continuity of measure

$$= \lim_{N \to \infty}P(\bigcap_{n \ge N} (X_n=0))$$

By monotone convergence theorem

$$= \sup_{N \ge 1}P(\bigcap_{n \ge N} (X_n=0))$$

By independence,

$$ = \sup_{N \ge 1} \prod_{n=N}^{\infty} P(X_n=0)$$

$$= \sup_{N \ge 1} \prod_{n=N}^{\infty} (1-p_n)$$

$$\to \sup_{N \ge 1} \prod_{n=N}^{\infty} (1-p_n) = 1$$

$$\to \forall \epsilon > 0, \exists m \ge 1 \text{s.t.} \prod_{n=m}^{\infty} (1-p_n) \in (1-\epsilon, 1]$$

Since $p_n < 1 \ \forall n \in \mathbb{N}$,

$$\forall \epsilon \in (0,1), \exists m \ge 1 \text{s.t.} \prod_{n=m}^{\infty} (1-p_n) \in (1-\epsilon, 1]$$

Hence $\forall \epsilon \in (0,1), \exists m \ge 1 \text{s.t.}$

$$\prod_{n=1}^\infty(1-p_n)=\prod_{n=1}^{m-1}(1-p_n)\prod_{n=m}^{\infty}(1-p_n) > 0 \ \because$$

  1. $\prod_{n=1}^{m-1}(1-p_n) > 0 \because p_n < 1 \ \forall n \ge 1$

  2. $\prod_{n=m}^{\infty}(1-p_n) > 0 \because \prod_{n=m}^{\infty} (1-p_n) \in (1-\epsilon, 1]$ QED


'Verified by Landon Carter'