Limiting distribution of the maximum of a null recurrent diffusion process (Q1932493)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Limiting distribution of the maximum of a null recurrent diffusion process
scientific article

    Statements

    Limiting distribution of the maximum of a null recurrent diffusion process (English)
    0 references
    0 references
    0 references
    18 January 2013
    0 references
    Consider a process \[ X(t),\;t\geq 0, \] on an open interval, with local generator \[ a(x)d^{2}/dx^{2}+b(x)d/dx=(d/dm(x))(d/ds(x)), \] whereby \(a(x)>0\), \(s\) is a scale transforming the interval into \(\mathbb{R}\), \(dm\) is the speed measure, \(X^{+}(t)=\max_{x\leq t}X(t)\), \(\xi(t)\) is another process with \[ \operatorname{P}(\xi (t_{i})\leq x_{i}, i=1,\dotsc,n)=\prod_{i=1}^{n}e^{-(t_{i}-t_{i-1})/x_{i}},\;\;0= t_{0}\leq t_{1}<\dotsb<t_{n},\;x_{i}<x_{i+1}, \] and a subordinator \(Z_{\alpha }(t)\) independent of \(\xi\), \(\operatorname{E}e^{-sZ_{\alpha}(t)}=e^{-ts^{\alpha }}\). Let \(\gamma >0\), \(\alpha =1/(\gamma +1)\), \(C_{\alpha }=\Gamma (1-\alpha)(\alpha (1-\alpha ))^{\alpha } /\Gamma (1+\alpha )\). Suppose that \[ \lim_{x\rightarrow -\infty }(-m(s^{-1}(x))|x|^{-\gamma })=c>0, \] \[ \lim_{x\rightarrow +\infty }(m(s^{-1}(x))x^{-\gamma })=0 \] (the exponent \(-1\) denotes the inverse functions). The first result states the convergence of the finite dimensional distributions of \(c^{\alpha }s(X^{+}(\lambda \cdot ))/\lambda^{\alpha }\) to those of \(\xi (Z_{\alpha }^{-1}(\cdot ))/C_{\alpha }\) for \(\lambda \rightarrow +\infty\). If also \[ \lim_{\lambda \rightarrow +\infty }(s^{-1}(\lambda x)-q(\lambda ))/\varphi (\lambda )=G(x)\;\text{ for}\;x>0 \] for some \(\varphi >0\), \(q\) and continuous \(G\), the second result is the same convergence for \((X^{+}(t)-q((t/c)^{\alpha }))/\varphi ((t/c)^{\alpha }))\) to \(G(\xi (1)Z_{\alpha }^{-1}(1)/C_{\alpha })\) for \(t\rightarrow +\infty\). An example with \(a=1/2\), \(b(x)=(p(x)-1)/2x\) follows, \(p=p_{-}\) on \((-\infty ,-1)\), \(p_{+}>0\) on \((1,+\infty )\), \(0<p_{-}<p_{+}<2\), and \(1\) on \([-1,1]\). The proofs rely on representations of all processes as functionals of one Brownian motion and on a theorem on the convergence of processes \(X(Y^{-1}(\cdot ))\) established in the appendix.
    0 references
    0 references
    diffusion
    0 references
    speed measure
    0 references
    scale function
    0 references
    inverse process
    0 references
    substitution
    0 references
    canonical extremal process
    0 references
    convergence of finite dimensional distributions
    0 references
    0 references