1
$\begingroup$

Is there a general (integral) solution to $du(t)= -a(t)u(t)\,dt +\sigma(t)\,dz$? Is the following $u(t)=e^{-\int_{t_0}^{t} \alpha(s) \, ds}u(t_0)+\int_{t_0}^t \sigma(v) e^{-\int_v^t a(s) \, ds} \, dz(v)$ correct (which I have seen claimed without a justification)? z(t) is the standard Wiener process. Is there a good reference for it?

$\endgroup$
  • $\begingroup$ What's $z$? ${}$ $\endgroup$ – Ian Feb 25 '17 at 4:26
  • $\begingroup$ z is the standard Wienner process. $\endgroup$ – Adam Feb 25 '17 at 20:57
1
$\begingroup$

About the differentiation formula of $u(t)$: Since \begin{align} u(t)&=e^{-\int_{t_0}^ta(s)\mathrm{d}s}u(t_0)+\int_{t_0}^t\sigma(v)e^{-\int_{v}^ta(s)\mathrm{d}s}\mathrm{d}Z(v)\\ &=e^{-\int_{t_0}^ta(s)\mathrm{d}s}\Bigl[u(t_0)+\int_{t_0}^t\sigma(v)e^{\int_{t_0}^va(s)\mathrm{d}s}\mathrm{d}Z(v)\Big]\stackrel{\text{def}}{=}x(t)y(t).\\ \mathrm{d}x(t)&=-a(t)e^{-\int_{t_0}^ta(s)\mathrm{d}s}\mathrm{d}t=-a(t)x(t)\mathrm{d}t,\\ \mathrm{d}y(t)&=\sigma(t)e^{\int_{t_0}^ta(s)\mathrm{d}s}\mathrm{d}Z(t)=\sigma(t)[x(t)]^{-1}\mathrm{d}Z(t). \end{align} Therefore, $$ du(t)=y(t)\mathrm{d}x(t)+x(t)\mathrm{d}y(t) =-a(t)u(t)\mathrm{d}t+\sigma(t)\mathrm{d}Z(t). $$

$\endgroup$
0
$\begingroup$

The $u(t)$ which you given is a solution of your equation. It is suffice using differentiation to verify it is a solution. One way to deduce the general solution is the `method of variation of constant' for ODE. The following book dealt with this topic: B. K. Oksendal, Stochastic Differential Equations, 6 Ed., Sringer Verlag, 2003.

$\endgroup$
  • $\begingroup$ Differentiating the integral formula, I get $du(t)=e^{\int_{t_0}^{t} a(s)ds}a(t) dt\cdot u(t_0)+\sigma(t)dz$. How do I reconcile it with $du=a(t)u(t)dt+\sigma(t)dz$? (I only see that when t is infinitesimally close to $t_0.$) $\endgroup$ – Adam Feb 26 '17 at 18:07

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.