Bootstrapping time series models*
An example of the time series data was seen in Chapter 1, Introduction to Ensemble Techniques, in the New Zealand Overseas dataset. See Chapter 10, Ensembling Survival Models, of Tattar et al. (2016). Time series is distinctive in that the observations are not stochastically independent of each other. For example, the maximum temperature of the day is very unlikely to be independent of the previous day's maximum temperature. However, we are likely to believe that the maximum temperature of a block of ten previous days is mostly independent of a ten-day block six months ago. Thus, the bootstrap method is modified to the block bootstrap method. The tsboot function from the boot package is useful to bootstrap time series data. The main structure of the tsboot function appears as follows:
tsboot(tseries, statistic, R, l = NULL, sim = "model",
endcorr = TRUE, n.sim = NROW(tseries), orig.t = TRUE,
ran.gen, ran.args = NULL, norm = TRUE, ...,
...
