Time series bootstrapping in R: How can I access each simulated path using tsbootstrap?

r, simulation, statistics, statistics-bootstrap, time-series

Solution

The `b` argument is the block length. `m` is the "block of blocks" argument for when you want to calculate statistics for each resampled series, rather than return each resampled series itself.

library(tseries)

# Simulate a time series
set.seed(1)
TS<-arima.sim(model=list(ar=c(.8,-.2)), n=20) 
plot(TS)

# 3 bootstrap samples with block size b=5
TSboot = tsbootstrap(TS, m=1, b=5, type="block", nb=3)

# Here are the individual bootstrapped series
TSboot

Time Series:
Start = 1 
End = 20 
Frequency = 1 
          [,1]        [,2]        [,3]
 1 -0.72571390  1.94273559  1.62729703
 2 -0.36463539  2.00048877  0.34495502
 3 -0.30236104  1.28640888 -2.26419528
...
18  0.96532247 -0.72571390 -0.36463539
19  1.59792898 -0.36463539 -0.30236104
20  1.67918002 -0.30236104 -1.63971414

plot(TSboot)

Problem

I want to perform a bootstrap analysis of a specific time series. I am using the function `tsbootstrap` of the package `tseries`. My problem: for values of m > 1, I cannot access each bootstrapped path individually (m: the length of the basic blocks in the block of blocks bootstrap, see `?tsbootstrap`) ``` library(tseries) set.seed(1) TS <- sample(1:20) tsbootstrap(TS,m=2, nb=1) ``` gives: ``` Error in tsbootstrap(TS, m = 2, nb = 1) : can only return bootstrap data for m = 1 ``` To my knowledge, the function can only compute some statistics (e.g. mean) over all of the simulated tranjectories, but I need each simulation itself. How can I come around this problem? (I am aware of the function `tsboot` of the package `boot`, but I was not able to operationalize the function yet)

Original source