Properties of a functional of trajectories which arises in studying the probabilities of large deviations of random walks (Q642075)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Properties of a functional of trajectories which arises in studying the probabilities of large deviations of random walks
scientific article

    Statements

    Properties of a functional of trajectories which arises in studying the probabilities of large deviations of random walks (English)
    0 references
    25 October 2011
    0 references
    Let \(\{\xi_n\}_{n\geq 1}\) be a sequence of independent copies of random vectors \(\xi:\Omega\to \mathbb{R}^d\) defined on a probability space \((\Omega,{\mathcal F},\operatorname{P})\). Let \[ \phi(\lambda)=\operatorname{E} e^{\langle\lambda,\xi\rangle},\;\Lambda(\alpha)= \sup_\lambda\{\langle\lambda, \alpha\rangle- \log\phi(\lambda)\},\;\alpha\in \mathbb{R}^d. \] Here, \(\langle\lambda, \alpha\rangle= \sum^d_{j=1} \lambda_{(j)}\alpha_{(j)}\), \(\lambda_{(j)}\) and \(\alpha_{(j)}\) are coordinates of \(\lambda\) and \(\alpha\). The local large deviation principle as applied to \(S_n= \sum^n_{j=1} \xi_j\), \(S_0= 0\), asserts that \[ \lim_{\varepsilon\to 0}\;\lim_{n\to\infty}\;{1\over n}\log\operatorname{P}\Biggl\{{S_n\over n}\in (\alpha)_\varepsilon\Biggr\}= -\Lambda(\alpha), \] where \((\alpha)_\varepsilon\) is the \(\varepsilon\)-neighborhood of \(\alpha\). Let us denote by \(C_\alpha[0,1]\) the space of all absolutely continuous functions \(f: [0,1]\to \mathbb{R}^d\) and by \(I(f)\) the Lebesgue integral \[ I(f)= \int^1_0 \Lambda(f')\,dt,\quad\text{where }f'(t)= (f_{(1)}',\dots, f_{(d)}'). \] Let us construct broken lines \(s+n =s_n(t)\), \(t\in [0,1]\), with their nodes \((k/n,S_n/n)\), \(k= 1,\dots, n\). Recall the known result that, if Cramer's conditions is fulfilled, i.e., \(\phi(\lambda)<\infty\) in a neighborhood of any point \(\mu\in\mathbb{R}^d\), then \[ \lim_{\varepsilon\to 0}\;\lim_{n\to\infty}\;{1\over n}\log\operatorname{P}\{s_n\in (f)_\varepsilon\}= \begin{cases} I(f),\quad &\text{if }f\in\mathbb{C}_a[0,1],\;f(0)= 0,\\ \infty,\quad &\text{in the remaining cases},\end{cases}\tag{1} \] where \((f)_\varepsilon\) is a \(\varepsilon\)-neighborhood of \(f\) in the uniform metric. The authors study (1) under weaker conditions, namely, Cramer's condition is fulfilled in a neighborhood of some special points \(\mu\in\mathbb{R}^d\). In the one-dimensional case, they extend (1) to a larger space \(C'[0,1]\subset C^1[0,1]\), \(C_a[0,1]\subset C'[0, 1]\) with integral \(J(f)\) instead of \(I(f)\) such that \(J(f)= I(f)\) for \(f\in C_a[0,1]\).
    0 references
    Cramér condition
    0 references
    deviation function
    0 references
    random walk
    0 references
    deviation functional
    0 references
    deviation integral
    0 references
    variation of a function
    0 references
    semi-additive function
    0 references
    Darboux integral
    0 references

    Identifiers