Kybernetika 38 no. 4, 389-404, 2002

Bootstrap in nonstationary autoregression

Zuzana Prášková


The first-order autoregression model with heteroskedastic innovations is considered and it is shown that the classical bootstrap procedure based on estimated residuals fails for the least-squares estimator of the autoregression coefficient. A different procedure called wild bootstrap, respectively its modification is considered and its consistency in the strong sense is established under very mild moment conditions.