Estimate Confidence Intervals for Ratios Using Beta Distribution

This function computes confidence intervals for ratios using the beta distribution, for each unique pair of numerator and denominator in the provided data. It supports both sequential and parallel computation, and allows customization of the ratio function, confidence level, and beta estimation parameters.

Usage

estimate_beta_df(
  data,
  fun = function(a, b) {
a/b * 100
 },
  D,
  V,
  js = 0,
  beta_min = 0,
  beta_max = 10,
  precision = 0.1,
  alpha = 0.05,
  posterior = FALSE,
  parallel = FALSE,
  max_cores = NULL
)

Arguments

data

A data frame or tibble with two columns: numerators and denominators.

fun

Function. Statistic calculation (default: a/b*100)

D

Integer. Maximum deviation

V

Numeric. Noise variance

js

Integer. Sensitivity threshold

beta_min

Numeric. Minimum of the search range for \(\beta\)

beta_max

Numeric. Maximum of the search range for \(\beta\)

precision

Numeric. Search step size

alpha

Numeric. Target error level

posterior

Logical. Use posterior approach? (default: FALSE)

parallel

Logical. If TRUE, computations are performed in parallel using available CPU cores. Default is FALSE.

max_cores

Integer or NULL. Maximum number of CPU cores to use for parallel computation. If NULL, uses all but one core.

Value

A tibble with columns:

A

Numerator value

B

Denominator value

R

Computed ratio using fun

alpha

Significance level used

beta

Estimated confidence interval or beta parameter

Details

The function removes duplicate rows from the input data before computation. Parallel computation is handled via the future and furrr packages.

Examples

if (FALSE) { # \dontrun{
nums <- c(20, 80)
denoms <- c(100, 200)
test <- data.frame(
  A = sort(rep(nums,length(denoms))),
  B = rep(denoms, each = length(nums))
)

fun = \(a,b){a/b * 100}
D = 5
V = 1

# A priori approach
res <- estimate_beta_df(test, fun, D, V)
# A posteriori approach
res_ap <- estimate_beta_df(test, fun, D, V, posterior = TRUE)
} # }