ABSTRACT
The bootstrap method is described by B. Efron and R. J. Tibshirani. It is a remarkably robust method of estimating variance statistics from the sample data itself. The bootstrap method involves repeated sampling of the data with replacement to estimate the variability associated with a statistic. To create a block bootstrap system for serially autocorrelated data, the first step is to create the blocks. The bootstrap errors are selected randomly in the same way as the original bootstrap values are selected, by sampling from the original data set with replacement, and then they are fitted into the parametric model of the random process. The block bootstrap is implemented by modification of the uncorrected bootstrap analogous to that of the previous section on time series. Code similar to that just displayed is used to construct a three-dimensional array based on the transpose of the data lattice, and the third dimension of the array is then sampled with replacement.
