A question regarding the analytic expression of an Ito integral

52 Views Asked by At

Recently I'm starting to learn Ito calculus from the book, Stochastic Process for Physicists by Kurt Jacobs. However, I have been puzzled by the following basic statement in pp.46 and pp.47 of the book for a long time.

Suppose we have the following integral:

\begin{equation} I = \int_0^{T} \left[\int_0^{T} f(s)dW(s) \right] g(t)dW(t) \end{equation}

The book states that the integral $I$ does not have an analytical expression. This is the point that confuses me a lot. In my understanding, if you write this integral as a limit of finite summations, we have:

\begin{equation} I = \lim_{N\to \infty} \sum_{i=1}^{N} \sum_{j=1}^{N} f_i g_j \Delta W_i \Delta W_j \end{equation}

Where $f_i = f(iT/N)$, $g_j = f(jT/N)$, $\Delta W_i = \Delta W(iT/N)$, $\Delta W_j = \Delta W(jT/N)$. You can separate the summation above as a product of two summations to get:

\begin{aligned} I &= \lim_{N\to \infty} \left(\sum_{i=1}^{N} f_i \Delta W_i \right) \left(\sum_{j=1}^{N} g_j \Delta W_j \right) \\ &= \lim_{N\to \infty} \left(\sum_{i=1}^{N} f_i \Delta W_i \right) \lim_{N\to \infty} \left(\sum_{j=1}^{N} g_j \Delta W_j \right) \\ &= F(T)G(T) \end{aligned}

Where $F(T)=\int_{0}^{T} f(t)dW(t)$ and $G(T)=\int_{0}^{T} g(s)dW(s)$. With this representation, we have expressed $I$ as the product of two random variables, $F(T)G(T)$. Since $F(T)$ and $G(T)$ are mutually dependent as they both reply on the same stochastic process $dW$, we can know $G(T)$ given $F(T)$. With that, we should be able to derive the distribution of $F(T)G(T)$.

So why is the integral $I$ claimed not to have an analytical expression? Could there be something wrong or missing in my derivation, or did I misunderstand the phrase "analytical expression"? If I misunderstood, what does it mean for an Ito integral to have "analytical expression"? I will greatly appreciate your guidance!