ABSTRACT

This talk aims to motivate and describe two formulations of empirical
likelihood for time series inference based on tapered data. One mpirical
likelihood is formed from tapered blocks of data in the time-domain, where data blocks help to capture the time dependence. The second empirical likelihood is based on the tapered periodogram (i.e., a data transformation) in the frequency-domain. Both approaches produce log-likelihoods with chi-squared limits for calibrating confidence regions.  Some simulations indicate that a data taper improves the coverage accuracy of empirical likelihood confidence intervals for time series parameters, such as means and
correlations.

Mon., October 8, 2007
4:00 p.m.
223 Weber