0
$\begingroup$

This is the buildup to establishing the Optional Sampling Theorem for continuous time. Let $\mathbb{F}$ denote a filtration and $(X_t)_{t\geq 0}$ be a right continuous supermartingale. Let $\sigma$ and $\tau$ be bounded stopping times, and define $\sigma_n=\frac{\lceil 2^{n}\sigma\rceil}{2^n}$, (likewise for $\tau_n$). We need to show that $\mathbb{E}[X_{\tau_n}|\mathcal{F}_{\sigma_m}]\to\mathbb{E}[X_{\tau_n}|\mathcal{F}_{\sigma}]$ a.s. as well as in $L^1$. So far, the only thing that is clear to me is that we should have $X_{\sigma_m}\to X_{\sigma}$ a.s throgh right continuity. Not quite sure how to proceed

$\endgroup$
0
$\begingroup$

The statement is a consquence of Lévy's backwards convergence theorem

Let $Y \in L^1$ be an integrable random variable and $\mathcal{G}_0 \supseteq \mathcal{G}_1 \supseteq \dots$ a decreasing sequence of $\sigma$-algebras. If we set $\mathcal{G}_{\infty} := \bigcap_{j \geq 1} \mathcal{G}_j$, then $$\mathbb{E}(Y \mid \mathcal{G}_j) \xrightarrow[]{j \to \infty} \mathbb{E}(Y \mid G_{\infty})$$ almost surely and in $L^1$.

In your setting, $\mathcal{G}_j := \mathcal{F}_{\sigma_j}$ and $Y=X_{\tau_n}$ for fixed $n$. Check that $(\mathcal{G}_j)_{j \geq 1}$ is a decreasing sequence of $\sigma$-algebras (here you have to use that $\sigma_{n+1} \leq \sigma_n$) and that $$\mathcal{F}_{\sigma} = \bigcap_{j \geq 1} \mathcal{F}_{\sigma_j}$$ (for this you have to use that $\sigma = \inf_j \sigma_j$). Applying Lévy's backward convergence theorem then proves the assertion.

| cite | improve this answer | |
$\endgroup$

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.