A0314
Title: A practical resampling-based approach to interval estimation for spectral densities
Authors: Daniel Nordman - Iowa State University (United States) [presenting]
Abstract: The spectral density function can play an important role in time series analysis, where nonparametric interval estimation of the spectral density becomes useful. Existing interval methods for spectral densities, including chi-square approximation and frequency domain bootstrap (FDB), can often be problematic in practice because intervals exhibit poor coverage. An alternative approach is presented that merges empirical likelihood (EL) and FDB. The idea is that EL provides a new statistic for setting intervals for spectral densities that can be effectively calibrated by bootstrap. The FDB-EL procedure is valid under mild conditions for application to a wide range of time processes. Numerical studies suggest that FDB-EL confidence intervals exhibit good performance compared to other methods. The confidence interval procedure is illustrated with an application to wind turbines.
