Аннотация:
Smooth bootstrap methods have received considerable attention in the case of independent identically distributed data, while smooth bootstrap methods for time series have received comparatively little. In a smooth bootstrap, the resampled values are perturbed by independent realizations from some kernel density. This has the effect of smoothing the empirical distribution from which the bootstrap samples are drawn, which improves the estimation of sampling distributions of statistics with distributions that depend on the population density. A smooth block bootstrap procedure is proposed and its consistency is established for a variety of statistical functionals on the marginal distribution of a stationary time series. (This is a joint work with Soumendra Lahiri and Dan Nordman)
Язык доклада: английский
|
