I was studying an example given in Variance of a time integral with respect to a Brownian Motion function

There we need to calculate the Variance of $I_t = \int\limits_{0}^{t} f \left(s \right) W_s ds$.

Then, it said that $E \left[ I_t^2 \right] = \int\limits_{0}^{t}\int\limits_{0}^{t} f \left(s \right) f \left(u \right) \min \left(s,u\right) dsdu$

Can someone please help to understand the detailed calculation on how it can be arrived?

It is also stated that $I_t$ follows a Normal distribution. What is the reason for that?


This uses the autocorrelation of the Weiner process (proved in this post), ${\mathbb E}[W_s W_u] = \min(s,u)$

From your expression, \begin{align} {\mathbb E}[I^2_t] &= {\mathbb E}[\int_0^t f(s)W_sds \int_0^t f(u)W_u du]\\ &= {\mathbb E}[\int_0^t \int_0^t f(s)W_s f(u)W_u ds du]\\ &= \int_0^t \int_0^t f(s) f(u) {\mathbb E}[W_s W_u] ds du \end{align} where we moved the expectation into the integral as the Weiner terms are the only non-deterministic things

Then we substitute in the autocorrelation, and bingo! \begin{align} {\mathbb E}[I^2_t] &= \int_0^t \int_0^t f(s) f(u) {\mathbb E}[W_s W_u] ds du\\ &= \int_0^t \int_0^t f(s) f(u) \min(s,u) ds du \end{align}

as required

  • $\begingroup$ Ahh and as for the normality of $I_t$ - remember that the sum of any number of normal variables is still normally distributed, even if they're correlated. An integration is the limit of a sum, so hopefully it's easy enough to believe that the sum of these brownians (weighted at each step by the local value of $f(s)$) must also be brownian $\endgroup$ – StackG Aug 4 '20 at 11:37

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.