Advanced Time Series Bootstrap Tools • tsbs

Overview

tsbs provides advanced bootstrap methods for time series data, with a focus on capturing complex temporal dependencies including regime-switching behavior and conditional heteroskedasticity. The package implements several bootstrap approaches ranging from classical block methods to sophisticated model-based resampling.

Installation

Install from GitHub:

# install.packages("devtools")
devtools::install_github("mahovo/tsbs")

Bootstrap Methods

The package supports the following bootstrap types via the main tsbs() function:

Method	`bs_type`	Description
Moving Block	`"moving"`	Fixed-length blocks sampled with replacement
Stationary Block	`"stationary"`	Random block lengths (geometric distribution)
Hidden Markov Model	`"hmm"`	Regime-switching with discrete states
MS-VAR	`"msvar"`	Markov-Switching Vector Autoregression
MS-VARMA-GARCH	`"ms_varma_garch"`	Markov-Switching VARMA with GARCH errors
Wild Bootstrap	`"wild"`	Residual-based with random sign flips

Quick Start

library(tsbs)

# Simple stationary block bootstrap
set.seed(123)
x <- arima.sim(n = 200, list(ar = 0.8))

result <- tsbs(
  x = as.matrix(x),
  bs_type = "stationary",
  block_length = 10,
  num_blocks = 20,
  num_boots = 100,
  func = mean
)

# Bootstrap distribution of the mean
hist(unlist(result$func_outs), main = "Bootstrap Distribution")
abline(v = result$func_out_means, col = "red", lwd = 2)

Documentation

Primary Documentation

The main entry point for the package is the tsbs() function, which provides comprehensive documentation including detailed specification examples:

file:///Users/mhvpbp13/git/tsbs/docs/reference/tsbs.html

In R with tsbs package installed and loaded:

?tsbs

Full reference manual:
https://mahovo.github.io/tsbs/

Vignettes

The package includes vignettes for specialized topics:

Vignette	Description
`vignette("portfolio-optimization-demo", package = "tsbs")`	Comprehensive demo of portfolio optimization w/ tsbs
`vignette("bootstrap_diagnostics", package = "tsbs")`	Diagnostic system for output bootstrap series
`vignette("ms-varma-garch_diagnostics.Rmd", package = "tsbs")`	Diagnostic system for monitoring EM convergence, parameter evolution, and numerical stability
`vignette("ms-varma-garch_inference", package = "tsbs")`	Statistical inference for DCC, CGARCH and GOGARCH parameters
`vignette("multivariate_garch_comparison", package = "tsbs")`	Diagnostic vignette comparing DCC, CGARCH and GOGARCH models

Demos

See a pre-rendered version of “portfolio-optimization.Rmd” :
https://github.com/mahovo/tsbs/blob/main/inst/demos/portfolio-optimization.md

library(tsbs)

# See what demos are available
list_package_demos()

# Get path to a demo  (choose demo name from list)
demo_path <- get_demo_path("portfolio-optimization")

# Render it (choose your output location)
demo_output_dir <- "demo-output"
rmarkdown::render(demo_path, output_dir = demo_output_dir)

# Or use in a pipeline
get_demo_path(demo_list[[1]]) |> 
  rmarkdown::render(output_dir = demo_output_dir)

Learning from Test Files

The package test files contain extensive examples of model specification and usage patterns. Examining these files can be a helpful way to understand how to specify models for different scenarios:

# Location of test files (after installation)
system.file("tests", package = "tsbs")

Key test files include:

test-ms_varma_garch_bs.R — Comprehensive examples of MS-VARMA-GARCH specifications
test-cgarch.R — Testing the CGARCH integration
test-gogarch.R — Testing the GOGARCH integration
test-blockBootstrap
test-hmm_bootstrap
test-wild_bootstrap

tsbs builds on the following packages for GARCH modeling:

tsgarch — Univariate GARCH models
tsmarch — Multivariate GARCH (DCC, Copula, GOGARCH)
MGARCH — Markov-Switching GARCH models

Key Features

Unified Interface: Single tsbs() function for all bootstrap methods
Flexible Function Application: Apply arbitrary functions to bootstrap replicates
Parallel Processing: Built-in support for parallel computation
Comprehensive Diagnostics: Monitor convergence and numerical stability for complex models

Dependencies

Core dependencies: - R (>= 4.0) - Rcpp - tsgarch - tsmarch

License

GPL (>= 3)

Citation

If you use tsbs in your research, please cite:

@software{tsbs,
 title = {tsbs: Time Series Bootstrap Methods},
 author = {Vognsen, Martin Hoshi},
 url = {https://github.com/mahovo/tsbs}
}

Contributing

Contributions are welcome! Please open an issue or submit a pull request on GitHub.

tsbs: Advanced Time Series Bootstrap Tools