Rice formula for processes with jumps and applications (Q2340036): Difference between revisions

Let \(T>0\), let \(C=\{C_t\}_{0\leq t\leq T}\) be an \(\mathbb{R}\)-valued process with continuously differentiable sample paths and let \(J=\{J_t\}_{0\leq t\leq T}\) be an \(\mathbb{R}\)-valued pure jump process admitting only finitely many jumps, independent from \(C\). In the paper under review, the authors study the number of crossings (and obtain Rice-type formulas for the mean number of continuous and discontinuous crossings) through a fixed level \(x\in\mathbb{R}\) by the process \(X:=C+J\). More precisely, by assuming that (i) \(C\) admits a density function \(p_C(t,y)\) which is jointly continuous for \(t\in[0,T]\) and \(y\) in a neighborhood of \(x\), (ii) for every \(s,t\in[0,T]\), the joint distribution of \((C_s,\dot{C}_t)\) admits a density \(p_{C,\dot{C}}(s,t,y,z)\) which is continuous in \(s\in[0,T]\) (\(t,y,z\) fixed) and in \(y\) at \(x\) (\(s,t,z\) fixed), (iii) for every \(t\in[0,T]\), there is a continuous version of \(\mathbb{E}[\dot{C}_t|X_t=y]\) in a neighborhood of \(x\) and (iv) \(\lim_{\delta\to 0}\sup_{0\leq s<t\leq T,\, |s-t|<\delta}|\dot{C}_t-\dot{C}_s|=0\) a.s., they show that the mean number of continuous and discontinuous crossings through \(x\) by \(X\) are given, respectively, by \[ \mathbb{E}[N^{c}_x]=\int_{[0,T]}\mathbb{E}[|\dot{C}_t||X_t=x]p_X(t,x)\, dt, \] \[ \mathbb{E}[N^{d}_x]=\int_{[0,T]\times\mathbb{R}}\mathbb{P}\left((X_{t^{-}}-x)(X_{t^{-}}+y-x)<0\right)\nu(dt,dy), \] where \(\nu(dt,dy)\) is the compensating measure of the counting measure generated by \(J\). Furthermore, using the above Rice-type formulas, the authors also discuss the mean number of continuous and discontinuous up-crossings through the level \(x\) by \(X\) (and obtain explicit expressions) for two examples: (i) \(C\) is a centered stationary Gaussian process with continuously differentiable sample paths (hence, it automatically satisfies the above assumptions) and \(J\) is a Poisson-auto-regressive process, and (ii) \(C\) is again a centered Gaussian stationary process with continuously differentiable sample paths and \(J\) is a compound Poisson process with standard Gaussian jumps. Finally, directly from these expressions they also conclude that, in the first example, the number of continuous up-crossings dominates over the number of discontinuous ones in mean, as the level goes to infinity (\(x\nearrow\infty\)). On the other hand, in the second example, the mean numbers of continuous and discontinuous up-crossings, as the level goes to infinity, are comparable.