ore
15:15
presso Enriques
A finite-variance stationary time displays long-range dependence
or long memory if its covariance function decays to zero like a power.
The decay should be slow enough so that the sum of the covariances
diverges. This happens, for example, if the corresponding spectral
density blows up at zero frequency like a power function.
These types of time series have been widely studied.
Normalized sums of such time series can converge to a Gaussian process,
typically fractional Brownian motion, but also to non-Gaussian processes,
which can be represented by multiple integrals. But unfortunately,
even the marginal distributions of these non-Gaussian processes
are not known explicitly. The simplest non-Gaussian member of this
family is the Rosenblatt process. It is represented by a double integral.
We shall study its marginal distributions and describe a way to obtain
them numerically. This is joint work with Mark Veillette.