Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let $(\Omega, \mathcal F , \mathbb P)$ be a probability space and let $\mathcal F = \{\mathcal F_t\}_{t\ge} $ a filtration. Let $W=\{W_t;t ≥ 0\}$ be a stochastic process adapted to $\mathcal F$. Suppose that W is a zero-mean Gaussian process with covariance function $$E(W_s W_t) = \min(s, t), \quad ∀s,t ∈ [0, ∞).$$ Show that W is a standard Wiener process.

In other words, I have to prove that the following conditions are satisfied:

  1. $W_0=0$
  2. $W_t-W_s$ is Gaussian with zero mean and $Var=|t-s|$
  3. increments of W corresponding to non-overlapping time intervals are independent
share|improve this question
2  
What have you tried? –  saz May 23 '13 at 16:30
    
I got the 3. relationship between covariance and indipendence here –  user79144 May 23 '13 at 16:55
    
this may be picky, but the filtration should be that of the process, not just one to which W is adapted. For example, choosing $\mathcal F = \lbrace \mathcal F_{\infty} \rbrace $ does not affect covariances but does affect whether you want to call it a wiener process. –  mike May 23 '13 at 17:12
    
Which point(s) is (are) eluding you, in the end? –  Did May 23 '13 at 20:52

1 Answer 1

Hints

  1. Choose $t=s=0$ in $\mathbb{E}(W_s \cdot W_t) = \min\{s,t\}$. (Note that $N(0,0)$ equals -by definition- $\delta_0$.)
  2. Let $s<t$. By assumption, the random vector $X:=(W_s,W_t)$ is Gaussian with mean $0$ and covariance matrix $$C:=\begin{pmatrix} s & s \\ s & t \end{pmatrix}$$ Since $X$ is Gaussian, we know that $\ell^T \cdot X = \ell_1 \cdot W_s+\ell_2 \cdot W_t$ is Gaussian for $\ell \in \mathbb{R}^2$. There are known formulas how to calculate the mean and variance of $\ell^T \cdot X$. Find a suitable $\ell$.
  3. Let $0:=t_0 < t_1 < \ldots < t_n$. Note that $$\Delta := \begin{pmatrix} W_{t_1}-W_{t_0} \\ \vdots \\ W_{t_n}-W_{t_{n-1}} \end{pmatrix} = \underbrace{\begin{pmatrix} -1 & 1 & 0 & 0 &\ldots & 0 \\ 0 &-1 & 1 & 0 & \ldots & 0 \\ \vdots & & \ddots & \ddots & & \\ 0 & 0 & 0 & \ldots & -1 & 1 \end{pmatrix}}_{=:M} \cdot \begin{pmatrix} W_{t_0} \\ \vdots \\ W_{t_n} \end{pmatrix}$$ Since $(W_{t_0},\ldots,W_{t_n})$ is Gaussian, by assumption, we conclude that $\Delta$ is Gaussian. Therefore, it suffices to show that the covariance matrix of $\Delta$ is a diagonal matrix. Again, there are known formulas how to calculate the covariance matrix of a linear transformation of a Gaussian random vector.

Remark In general, one requires that the paths $t \mapsto W(t,w)$ are continuous for almost all $w$. The theorem of Kolmogorov-Chentsov shows that, under the given assumptions, there exist always a modification $(\tilde{W}_t)_t$ of the process $(W_t)_t$ such that the sample paths $t \mapsto \tilde{W}(t,w)$ are continuous almost surely.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.