Ornstein-Uhlenbeck process

From Wikipedia, the free encyclopedia

In mathematics, the Ornstein-Uhlenbeck process (named after Leonard Salomon Ornstein and George Eugene Uhlenbeck), also known as the mean-reverting process, is a stochastic process given by the following stochastic differential equation

$dr_t = -\theta (r_t-\mu)\,dt + \sigma\, dW_t,\,$

where θ, μ and σ are parameters and W_t denotes the Wiener process.

three sample paths of different OU-processes with θ=1, μ=1.2, σ=0.3:
navy: initial value a=0 (a.s.)
olive: initial value a=2 (a.s.)
red: initial value normally distributed so that the process has invariant measure

[edit] Solution

This equation is solved by variation of parameters. Apply Itō's lemma to the function $f (r t, t) = r t e θ t$ to get

$df(r_t,t) = \theta r_t e^{\theta t}\, dt + e^{\theta t}\, dr_t\,$

$= e^{\theta t}\theta \mu \, dt + \sigma e^{\theta t}\, dW_t. \,$

Integrating from 0 to t we get

$r_t e^{\theta t} = r_0 + \int_0^t e^{\theta s}\theta \mu \, ds + \int_0^t \sigma e^{\theta s}\, dW_s \,$

whereupon we see

$r_t = r_0 e^{-\theta t} + \mu(1-e^{-\theta t}) + \int_0^t \sigma e^{\theta (s-t)}\, dW_s. \,$

Thus, the first moment is given by (assuming that $r 0$ is a constant),

E (r t) = r 0 e - θ t + μ(1 - e - θ t).

Denote $s \wedge t = \min(s,t)$ we can use the Itō isometry to calculate the covariance function by

$\operatorname{cov}(r_s,r_t)= E[(r_s - E[r_s])(r_t - E[r_t])]$

$= E[\int_0^s \sigma e^{\theta (u-s)}\, dW_u \int_0^t \sigma e^{\theta (v-t)}\, dW_v ]$

$= \sigma^2 e^{-\theta (s+t)}E[\int_0^s e^{\theta u}\, dW_u \int_0^t e^{\theta v}\, dW_v ]$

$= \frac{\sigma^2}{2\theta} \, e^{-\theta (s+t)}(e^{2\theta (s \wedge t)}-1).\,$

It is also possible (and often convenient) to represent $r t$ (unconditionally) as a scaled time-transformed Wiener process:

$r_t=\mu+{\sigma\over\sqrt{2\theta}}W(e^{2\theta t})e^{-\theta t}$

or conditionally (given $r 0$ ) as

$r_t=r_0 e^{-\theta t} +\mu (1-e^{-\theta t})+ {\sigma\over\sqrt{2\theta}}W(e^{2\theta t}-1)e^{-\theta t}.$

The Ornstein-Uhlenbeck process (an example of a Gaussian process that has a bounded variance) admits a stationary probability distribution, in contrast to the Wiener process.

The time integral of this process can be used to generate noise with a 1/f power spectrum.