I'm not taking an official class (actuarial exams), some fellow "students" created a question (forum discussion), considering the integral in title. This is my attempt at a solution with no real justifications

\begin{align} \int_0^tsZ_s \, ds & = \int_0^ts\lim_{n\to\infty}\sum_{i=1}^nY(ih)\sqrt h\,ds\\ & = \lim_{n\to\infty}\frac{1}{\sqrt n}\sum_{i=1}^nY(ih)\int_0^ts^{3/2} \, ds\\ & = \lim_{n\to\infty}\frac{1}{\sqrt n}\sum_{i=1}^nY(ih)\frac{2}{5}t^{5/2}\\ & = \frac{2}{5}t^2Z_t. \end{align}

I have strong doubts that this is correct. Or if it is correct, how should I confirm? Using Ito's lemma? I am using the simple representation of Wiener motion,

$$Z_t=\lim_{n\to\infty}\sum_{i=1}^nY(ih)\sqrt h,$$

with step size $h=t/n$, where $Y(ih)=\pm1$ each with probability $\frac{1}{2}$ (transformation of a Bernoulli RV).

I have taken the series of undergraduate real analysis courses, so one idea that pops in my head is that $Z$ is not well-behaved enough to interchange objects such as limits, sums and integrals. I would find it interesting to know which theorems' conditions I am not meeting, from various courses along the sequence of courses ascending to a course in stochastic integration.


The fastest way to verify your formula is to apply Ito's lemma: $$ \mathrm d(\frac25t^2Z_t) = \frac45tZ_t + \frac25t^2\mathrm dZ_t\neq tZ_t $$ so the answer is incorrect. Another point is that the representation via $Y$ amy not be applicable in case of computing integrals. Also you have to remember that in your case you have different representations for each $Z_s$ where $s$ runs in $[0,t]$ - but those representations must be clearly dependent. In particular, it is unclear to me how did you get $s^{\frac32}$ and took $Y$ outside of the integral.

In case $Z_t$ is a standard Brownian motion, sometimes it hepls to use the integration by parts: $$ \mathrm d(f_tZ_t) = f'_tZ_t\mathrm d t+f_t\mathrm dZ_t \implies f'_tZ_t\mathrm = \mathrm d(f_tZ_t) - f_t\mathrm dZ_t $$ which holds for any deterministic $C^1$ function $f_t$. Now, in your case $f'_t = t$ and so $$ \int_0^t sZ_s\mathrm ds = \frac12 t^2Z_t - \frac12\int_0^t s^2\mathrm dZ_s $$ which does not seem to be a much of simplification, though. In the end, it does not seem to me that the original integral can be simplified.


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.