Nonstationary self-similar Gaussian processes as scaling limits of power-law shot noise processes and generalizations of fractional Brownian motion

Abstract

We study shot noise processes with Poisson arrivals and nonstationary noises. The noises are conditionally independent given the arrival times, but the distribution of each noise does depend on its arrival time. We establish scaling limits for such shot noise processes in two situations: (a) the conditional variance functions of the noises have a power law and (b) the conditional noise distributions are piecewise. In both cases, the limit processes are self-similar Gaussian with nonstationary increments. Motivated by these processes, we introduce new classes of self-similar Gaussian processes with nonstationary increments, via the time-domain integral representation, which are natural generalizations of fractional Brownian motions.