AdaptiveBoxplot (version 0.1.1)

bh_boxplot: False Discovery Rate (FDR) Boxplot

Description

Generates a boxplot where whisker lengths are determined by the Benjamini-Hochberg procedure to control the False Discovery Rate (FDR), making the outlier detection rule adaptive to sample size and data characteristics.

Usage

bh_boxplot(data, alpha = 0.01, group_col = NULL, value_col = NULL, ...)

Value

A plot is drawn on the current graphics device.

Arguments

data

A numeric vector for a single boxplot, or a data frame for grouped boxplots.

alpha

The target FDR level. Defaults to 0.01.

group_col

A string specifying the name of the grouping column in `data`.

value_col

A string specifying the name of the value column in `data`.

...

Additional arguments passed to the base boxplot function.

Details

This function is a graphical implementation of the p-value pipeline proposed by Gang, Lin, and Tong (2025). It uses robust estimators for the mean and standard deviation based on quartiles to calculate p-values for each observation, then applies the Benjamini-Hochberg (BH) procedure to determine an adaptive p-value threshold for outlier detection. Outliers are points falling beyond the fences defined by this threshold.

References

Gang, B., Lin, H., & Tong, T. (2025). Unifying Boxplots: A Multiple Testing Perspective.

See Also

holm_boxplot

Examples

# Single group example
set.seed(123)
data_single <- c(rnorm(50), 10, 12)
bh_boxplot(data_single, alpha = 0.05, main = "FDR Boxplot (Single Group)")

# Grouped data example
data_grouped <- data.frame(
  Category = rep(c("A", "B"), each = 100),
  Value = c(rnorm(100), rnorm(100, mean = 2, sd = 1.5))
)
bh_boxplot(data_grouped, group_col = "Category", value_col = "Value")