Let's say, that $B_t$, $t\geq0$ is standard Brownian motion (Wiener process). Let's define process $$X_t=B_t+B_{t^2}\text{, }t\geq0$$

I need to find its variance, covariance, find out if it's Gaussian and its stationarity.

I started from variance:

1) $EX_t=E(B_t+B_{t^2})=EB_t+EB_{t^2}=0$ ?

2) $cov(X_s,X_t)=E(X_sX_t)=E(X_s[X_t-X_s+X_s])=E(X_sX_t-X_{s^2}+X_{s^2})=\ldots$

any ideas how to show those things? How I can interpret sum of two Brownian motions?

Thanks for suggestions



  • You have correctly computed $E X_t$.

  • Recall that $\mathrm{Cov}(B_t,B_s) = t \wedge s$, and that $\mathrm{Cov}(\cdot,\cdot)$ is linear in both components.

  • Your answer for the variance and distributions will depend on whether $t<1$, $t=1$, or $t>1$.

  • Recall that Brownian motion has multivariate normal finite distributions. In particular $(B_t,B_{t^2})$ is bivariate normal (you compute the mean and covariance matrix (easy))

  • If $N$ is multivariate normal and $A$ is an affine transformation (like $(x,y) \to x + y$ cough cough)), then $AN$ is multivariate normal.

  • The sum of (jointly) stationary processes is stationary, can you show this?

  • $\begingroup$ Got it! Thanks : ) $\endgroup$ – atomoutside Jun 13 '13 at 19:03
  • $\begingroup$ "The sum of (jointly) stationary processes is stationary" is true but slightly misleading since it does not apply to the present case. $\endgroup$ – Did Jun 14 '13 at 6:00

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.