I'm following a solution of an SDE from here


Start with the SDE $$ dX_t = \delta dt + 2\sqrt{X_t} dW_t $$ consider a deterministic time change $$ \tau = \frac{\sigma^2}{2\nu(2-\delta)}\left(1-\exp\left(-\frac{2\nu t}{2-\delta}\right)\right) $$ the process $Y_t$ is defined as $$ Y_t = exp(\nu t) X_{\tau}^{1-\delta/2} $$ then, using Ito's lemma, we get $$ dY_t = \nu Y_t dt + \sigma Y_t^{\frac{1-\delta}{2-\delta}} dW_t $$

I'd like to understand how the Ito's lemma is applied here. The problem that I have is that when you write it, the $Y$ is derivated with respect to the $X$, but in the definition of $Y_t$ is used $\tau$, not $t$, in $X$, so there's a mixture between different times. How is this handled?


The time change described in the question may be handled as follows. Recall that if $W(t)$ is a standard Brownian motion then $$ W(\tau(b))-W(\tau(a)) $$ has the same distribution as $$ \int_a^b \sqrt{\tau'(t)} dW(t) $$ where $a \le b$. Therefore the time-changed process $\tilde X_t = X_{\tau(t)}$ satisfies: $$ d \tilde X_{t} = \delta \tau'(t) dt + 2 \sqrt{ \tilde X_{t} } \sqrt{\tau'(t)} dW(t) \;. $$ Now apply Ito's chain rule to $Y_t = \exp(\nu t) \tilde X_t^{1 - \delta/2}$ to obtain $$ d Y_t = \nu Y_t dt + \exp(\nu t) (1-\frac{\delta}{2}) \tilde X_{t}^{-\delta/2} d \tilde X_{t} +\underbrace{\exp(\nu t) \left( 2 ( -\frac{\delta}{2} ) (1-\frac{\delta}{2}) \tau'(t) \tilde X_{t}^{-\delta/2} \right) dt}_{\text{Ito correction to standard chain rule}} \;. $$ Eliminate $\tilde X(t)$ to obtain the desired SDE for $Y_t$.

  • $\begingroup$ great, I never saw a change in the time coordinates until now actually. Do you have a reference where this is considered explicitly? $\endgroup$ – chuse Apr 15 '15 at 13:07
  • $\begingroup$ Check Oksendal's book, SDE: An introduction with applications. $\endgroup$ – nabla Apr 17 '16 at 11:46

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.