Bootstrap-Based Statistical Inference for Dependent Data
Conference
Category: Bernoulli Society for Mathematical Statistics and Probability (BS)
Proposal Description
When the distribution of a statistic is not accessible, statistical inference is often based on asymptotic results. However, the small sample performance of confidence sets based on asymptotic critical values can be very poor. Even more the asymptotics often depend on unknown parameters and are therefore inaccessible. The bootstrap offers a convenient way to circumvent these difficulties. This session provides an overview of recent variants the bootstrap for time series with a complex dependence structure. The topics of the talks range from theoretical results on the validity to applications e.g. in econometrics.
Submissions
- A distance covariance based test for independence of long-range dependent time series
- Bootstrap convergence rates for the maximum of an increasing number of autocovariances and autocorrelations under strict stationarity
- Optimal choice of bootstrap block length for periodically correlated time series