1
$\begingroup$

Prove the stochastic process $Y = (Y_n)_{n \geq 0}$ is a martingale w/rt the filtration $\mathscr{F} = (\mathscr{F_n})_{n \geq 0}$, where $\mathscr{F_n} = \mathscr{F_n}^{Y} \doteq \sigma(Y_0, Y_1, ..., Y_n)$ where $Y$ is given by:

$Y_0 = 1$

and for $n \geq 1$

$Y_{n+1} = 2Y_n$ w/ prob 1/2

$Y_{n+1} = 0$ w/ prob 1/2

  1. $\sigma(Y_n) \subseteq \mathscr{F_n}$, obviously

  2. $Y_n$ is integrable because it is bounded: $Y_n \leq 2^n$

  3. $E[Y_{n+1} | \mathscr{F_n}] = Y_n$


What I tried doing:

Let $A_n = \{ \omega \in \Omega | Y_{n+1}(\omega) = 2 Y_n(\omega) \} \in \mathscr{F}$. Note that $P(A_n) = 1/2$.

$Y_n$ is a product of indicator functions namely: $Y_n = 2^n \prod_{0 \leq i \leq n-1} 1_{A_i}$

since

$Y_0 = 1$

$Y_1 = 2 Y_0 \times 1_{A_0} + 0 \times 1_{A_0^c} = 2^1 \times 1_{A_0}$

$Y_2 = 2 Y_1 \times 1_{A_1} + 0 \times 1_{A_1^c} = 2^2 \times 1_{A_0} \times 1_{A_1}$

and so on.

Now $E[Y_{n+1} | \mathscr{F_n}] = E[2^{n+1} \prod_{0 \leq i \leq n} 1_{A_i} | \mathscr{F_n}]$

$= 2^{n+1} \prod_{0 \leq i \leq n-1} 1_{A_i} E[1_{A_{n}} | \mathscr{F_n}]$

$= 2^{n+1} \prod_{0 \leq i \leq n-1} 1_{A_i} P[{A_{n+1}} | \mathscr{F_n}]$

$= 2^n \prod_{0 \leq i \leq n-1} 1_{A_i} = Y_n$.

Is that correct?

I seemed to assume $\sigma(1_{A_{n}}) = \sigma(A_n)$ and $\mathscr{F_n}$ are independent. Are they?

What is $\sigma(Y_n)$ in terms of $A_n$ anyway? I guess that:

$\sigma(Y_0) = \{ \emptyset, \Omega \}$

$\sigma(Y_1) = \sigma(A_0) \because Y_1 = 2 \times 1_{A_0}$

$\sigma(Y_2)$...$\subseteq \sigma(\sigma(A_0) \cup \sigma(A_1)) = \sigma(A_0, A_1)$

$\endgroup$
  • 1
    $\begingroup$ If $Y_n=2\prod _ {i\leq n-1} 1_{A_i}$, then $|Y_n| \leq 2$ for all $n$, but $Y_2 = 4$ with probability 1/4... $\endgroup$ – ekvall Sep 8 '15 at 1:20
  • $\begingroup$ @Student001 Oh, 2 should have an exponent! Thanks! $\endgroup$ – BCLC Sep 8 '15 at 1:26
0
$\begingroup$

Given a filtered probability space $(\Omega, \mathscr{F}, \{\mathscr{F_n}\}, \mathbb{P})$ where $\mathscr{F_n} = \mathscr{F_n}^{X} \doteq \sigma(X_0, X_1, \ldots, X_n)$, show that $X = (X_n)_{n \geq 0}$ is a $(\mathscr{F}_n^X, \mathbb{P})$-martingale where $X$ is given by:

$X_{n+1} = 2X_n$ w/ prob 1/2

$X_{n+1} = 0$ w/ prob 1/2

and $X_0 = 1$.

Define the iid random variables $V_0 = 1$,

$V_1, V_2, \ldots \sim P(V_i = 0) = P(V_i = 2) = 1/2$. Then, $X_n = \prod_{i=0}^{n} V_i$.

  1. $X_n$'s are bounded and hence integrable.
  2. $X_n$'s are adapted to their natural filtration.
  3. $E[X_n \mid \mathscr{F_m}] = X_m$

\begin{align} \text{LHS} & = E\left[\prod_{i=0}^{n} V_i \mid \mathscr{F_m}\right] \\ & = E\left[\prod_{i=0}^{n} V_i \mid \mathscr{F_m}\right] \\ & = E\left[\prod_{i=0}^{m} V_i \prod_{i=m+1}^{n} V_i\mid \mathscr{F_m}\right] \\ & = E\left[X_m \prod_{i=m+1}^{n} V_i\mid \mathscr{F_m}\right] \\ & = X_m E\left[\prod_{i=m+1}^{n} V_i\mid \mathscr{F_m}\right] \\ & = X_m E\left[\prod_{i=m+1}^{n} V_i\right] \tag{*} \\ & = X_m \prod_{i=m+1}^{n} E\left[V_i\right] \text{ by the independence of the $V_i$'s} \\ & = X_m \prod_{i=m+1}^{n} E\left[V_i\right] \\ & = X_m \prod_{i=m+1}^{n} (1) \\ & = \text{RHS} \quad \text{QED} \end{align}

(*)

$\mathscr G_m = \sigma(V_1,\ldots,V_m) \supset \mathscr F_m$

Being independent of $\mathscr G_m$ implies being independent of $\mathscr F_m$

$\endgroup$

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.