In the Stochastic analysis course we encountered the following integral

$\int_0^\infty H^2_sd[M,M]_s$,

where $H_s$ is a predictable process, $M_s$ is a uniformly integrable martingale in $L^2$, $[M,M]_s$ is the quadratic variation.

We were working on the measure space $(\mathbb{R}^+ \times \Omega, \mathcal{B}(\mathbb{R}^+) \otimes \mathcal{F}) $. As far as I understand, we consider the integral $w$-wise ($\omega \in \Omega$) as a Lebesgue–Stieltjes integral, with a stochastic process being a function $s \mapsto X_s(\omega), \ X_s(w): \mathbb{R}^+ \rightarrow \mathbb{R}$ for any $\omega$.

The question is: how the integral is defined (taking into account we are in $\mathbb{R}^+$ and the integral is to $\infty$)?

My guess was: we define it via a limit as $s \rightarrow \infty$ of the integral from 0 to $s$. However, my professor declined it (saying "there is no problem here as with a Riemann integral").


Let $\omega\in\Omega$ be fixed and let $F\!:\mathbb{R}_+\to \mathbb{R}_+$ be defined as $F(s)=[M,M]_s(\omega)$. Then $F$ is a non-negative, right-continuous and non-decreasing function with $F(0)=0$, and hence $$ \mu_F((a,b])=F(b)-F(a),\quad0\leq a<b<\infty $$ defines a Borel-measure on $(\mathbb{R}_+,\mathcal{B}(\mathbb{R}_+))$. If $g\!:\mathbb{R}_+\to \mathbb{R}_+$ is a measurable function, then integrating with respect to $F$ is just ordinary Lebesgue-integration with respect to $\mu_F$, i.e. $$ \int_0^\infty g(s)\,\mathrm dF(s):=\int_{\mathbb{R}_+}g(s)\mu_F(\mathrm ds).\tag{1} $$ Now, the integral $$ \int_0^\infty H_s^2\mathrm d[M,M]_s $$ is just the $\omega$-wise construction of the integral in $(1)$.


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.