Abstrakt
Various bootstrap procedures (parametric, wild and modified wild bootstrap)to approximate the distribution of the LSE of the autoregression coefficient in an AR(1) process with heteroscedastic innnovations are considered and their consistency is studied.