symbootpkg

The goal of symbootpkg is to provide flexible tools to compute bootstrap-t, equal-tailed, and symmetric confidence intervals for statistical analysis. It incorporates methods for plugin and nested bootstrap standard error estimations, with options for parallelization to improve efficiency.

Features

• Bootstrap-t Confidence Intervals: Compute confidence intervals using the bootstrap-t method with options for plugin and nested standard error estimates.
• Equal-Tailed and Symmetric Intervals: Includes methods for both equal-tailed and symmetric bootstrap confidence intervals.
• Flexible Standard Error Functions: Supports user-specified sdfun functions or nested bootstrap for standard error estimation.
• Parallel Processing: Optional parallelization to leverage multiple cores for faster computation.

Installation

You can install the development version of symbootpkg from GitHub with:

# install.packages("devtools")
devtools::install_github("blessingofori-atta/symbootpkg")

Usage

Basic Example

Here’s a basic example of how to use the bootstrap and bootstrap_t functions:

library(symbootpkg)
set.seed(123)

# Define a statistic function
beta <- function(dat) {
  mod <- lm(Sepal.Length ~ Sepal.Width, data = dat)
  bb <- mod$coefficients[2]
  return(bb)
}

# Perform bootstrap resampling
boot_obj <- bootstrap(
  data = iris,
  statistic = beta,
  nboot = 1000,
  return_samples = TRUE
)

# Compute symmetic bootstrap-t confidence intervals
ci_result <- bootstrap_t(
  boot_obj = boot_obj,
  Bsd = 25,
  method = "bootsym-nested",
  alpha = 0.05
)

# Print results
print(ci_result)
#> 
#> Bootstrap-t Results
#> =====================
#> Method:          bootsym-nested 
#> Alpha:           0.05 
#> Theta (Observed): -0.2233611 
#> Standard Error:  0.1422654 
#> Bias:            -0.003427493 
#> Confidence Interval:
#>   Lower:  -0.519971 
#>   Upper:  0.07324885

Parallel Processing Example

You can enable parallel processing to speed up the computation:

ci_result_parallel <- bootstrap_t(
  boot_obj = boot_obj,
  Bsd = 25,
  method = "bootsym-nested",
  parallel = TRUE,
  cores = 2,
  alpha = 0.05
)

# Print results
print(ci_result_parallel)
#> 
#> Bootstrap-t Results
#> =====================
#> Method:          bootsym-nested 
#> Alpha:           0.05 
#> Theta (Observed): -0.2233611 
#> Standard Error:  0.1422654 
#> Bias:            -0.003427493 
#> Confidence Interval:
#>   Lower:  -0.5131957 
#>   Upper:  0.06647356

Contribution

Contributions are welcome! Please open an issues or submit a pull request to improve the package.

License

This package is licensed under the MIT License. See the LICENSE file for details.

References

1. Efron, B., & Tibshirani, R. J. (1993). An introduction to the bootstrap. New York:Chapman & Hall.
2. Hall, P. (1988). On Symmetric Bootstrap Confidence Intervals. Journal of the Royal Statistical Society. Series B (Methodological), 50(1), 35–45. http://www.jstor.org/stable/2345806.
3. Elias, C. J. (2015). Percentile and percentile- t bootstrap confidence intervals: A practical comparison. Journal of Econometric Methods, 4(1), 153–161. doi:10.1515/jem-2013-0015
4. http://users.stat.umn.edu/~helwig/notes/npboot-notes.html

---