kernel.bandwidth: Kernel bandwidth estimation by median heuristic method with bootstrap CI

Description

Computes the Gaussian-kernel (Radial Basis Function) bandwidth \(\sigma\) using the median-heuristic (inverse of the median of squared Euclidean distances), and returns a bootstrap confidence interval. This assumes the kernel is parameterized as \(k(x_i, x_j) = \exp(-\sigma ||x_i - x_j||^2)\).

Usage

kernel.bandwidth(
  comm,
  scale_data = FALSE,
  ci_type = "bca",
  runs = 1000,
  conf = 0.95,
  cores = 1
)

Value

An object of class sigma_bootstrap_result containing:

sigma_original: Point estimate from the full data.
sigma_bootstrap: Bootstrap mean estimate.
bootstrap_se: Bootstrap standard error.
ci: Confidence interval object from boot.ci.
bootstrap_distribution: Vector of bootstrap values.
boot_object: The full boot object from boot().

Arguments

comm: A numeric matrix or data frame.
scale_data: Logical; if TRUE, variables are centered and scaled before distances are computed.
ci_type: Character; passed to boot.ci. One of "norm", "basic", "stud", "perc", or "bca".
runs: Integer; number of bootstrap resamples.
conf: Confidence level for the CI (between 0 and 1).
cores: Number of cores to be used in parallel processing. If = 0 all available cores are used.

References

Carvalho, J.C. & Cardoso, P. (2025) Quantifying species distribution within the functional space.

Examples

Run this code

set.seed(1)
comm <- matrix(rexp(200, rate = 0.1), ncol = 4)
res <- kernel.bandwidth(comm, runs = 99)
print(res)