Abstract:
|
Bootstrapping, a method of resampling with replacement, has simplified the estimation of standard errors and confidence intervals in a variety of intricate applications. Regarding time series data, a variety of bootstrapping methods have been proposed, such as the block and stationary bootstrap, the parametric bootstrap, the semi-parametric sieve and Cholesky bootstrap. The latter is a model free technique based on properties of the autocovariance structure. We study the sensitivity of the methods in their application for multivariate time series and apply methodology originally designed for high-dimensional multivariate analysis. A simulation study is conducted to compare the efficiency of these bootstrapping techniques.
|