Strong invariance principle for singular diffusions. (Q2574548)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Strong invariance principle for singular diffusions. |
scientific article |
Statements
Strong invariance principle for singular diffusions. (English)
0 references
29 November 2005
0 references
Let us consider a stochastic differential equation \[ d\xi (t) = b(\xi (t))\,dt + \sigma (\xi (t))\,dW(t) \tag{1} \] in \(\mathbb R^ {d}\), where \(W\) is a standard \(m\)-dimensional Wiener process and the coefficients \(b: \mathbb R^ {d}\to \mathbb R^ {d}\), \(\sigma : \mathbb R^ {d}\to \mathbb R^ {d}\otimes \mathbb R^ {m}\) are globally Lipschitz continuous. Set \(a(x,y) = (\sigma (x)- \sigma (y))(\sigma (x)-\sigma (y))^ *\) and suppose that the following Lyapunov condition is satisfied: \(2\langle b(x)-b(y), D(x-y)\rangle + \operatorname {Tr}(a(x,y)D) - 2\langle x-y,D(x-y)\rangle ^ {-1} \langle x-y, Da(x,y)D(x-y)\rangle \leq -\beta | x-y| ^ 2\) for some symmetric positive definite matrix \(D\), a sufficiently large constant \(\beta \) and all \(x,y\in \mathbb R^ {d}\), \(x\neq y\). These hypotheses do not imply that the matrix \(\sigma (x) \sigma (x)^ *\) is regular for all \(x\), nevertheless, it was proven by \textit{G. K. Basak} [J.\ Multivariate Anal.\ 39, 44--59 (1991; Zbl 0746.60076)] that there exists a unique invariant probability measure \(\mu \) for (1) and any solution to (1) converges in law to \(\mu \) as \(t\to \infty \). A multivariate strong invariance principle with good rate is established. Let \(f: \mathbb R^ {d}\to \mathbb R^ {N}\) satisfy a suitable Hölder continuity type condition, \(\mu (f)=0\). Let \(\xi \) be either a solution to (1) with a deterministic initial condition, or the strictly stationary solution to (1). Then there exist an \(N\)-dimensional Wiener process \(B\) and \(\eta >0\) such that \(| \int ^ {t}_ {0} f(\xi (s))\,ds - B(t)| = O(t^ {1/2-\eta })\) as \(t\to \infty \) almost surely. The covariance matrix \(G\) of \(B\) is given by \[ G_ {ij} = \int ^ \infty _ 0\mathbf E[f_ {i}(\xi (0))f_ {j} (\xi (t))]\,dt + \int ^ \infty _ 0\mathbf E[f_ {j}(\xi (0)) f_ {i}(\xi (t))]\,dt. \] Consequently, known asymptotic properties of the process \(B\) may be applied to the process \((\int ^ {t}_ 0 f(\xi (s))\, ds\), \(t\geq 0)\).
0 references
stochastic differential equations
0 references
invariant measure
0 references
0 references
0 references
0 references