Chauvenet-Type Boxplot.
ChauBoxplot
ChauBoxplot is an R package designed to create an improved version of the boxplot. This package implements a new fence coefficient proposed by Lin et al. (2025), replacing the traditional fence coefficient $k=1.5$ in Tukey's boxplot. The new coefficient $k=k_n^{Chau}$ is calculated based on Chauvenet's criterion, which is given in formula (9) in their paper.
Features
- chau_boxplot(): Generates Chauvenet-type boxplots in
base R. Its usage is similar to boxplot(), but it employs an updated fence coefficient for more robust outlier detection. - geom_chau_boxplot(): Creates Chauvenet-type boxplots in
ggplot2, functioning similarly to geom_boxplot().
Installation
To install the ChauBoxplot package from CRAN, please use the following command in R:
install.packages("ChauBoxplot")
To install the ChauBoxplot package from GitHub, please use the following commands in R:
library(devtools)
install_github("tiejuntong/ChauBoxplot")
Documentation
For detailed documentation and usage examples, please also visit the package website at https://github.com/tiejuntong/ChauBoxplot/.
Usage
Below is a real example with R code of how to create a Chauvenet-type boxplot for the pay adjustment rates of senior civil servants in Hong Kong.
R code in base R
library(ChauBoxplot)
rate.senior <- c(4.96, 6.30, -5.38, 1.60, 7.24, 5.26, 2.55, 5.96, 3.96, 4.19, 1.88, 4.06, 4.75, 0, 0, 2.5, 2.87, 3.00)/100 chau_boxplot(rate.senior)
R code in ggplot2
library(ggplot2)
library(ChauBoxplot)
rate.senior <- c(4.96, 6.30, -5.38, 1.60, 7.24, 5.26, 2.55, 5.96, 3.96, 4.19, 1.88, 4.06, 4.75, 0, 0, 2.5, 2.87, 3.00)/100
year <- 2007:2024
data.senior <- data.frame(x=year, y=rate.senior)
C.boxplot.senior <-
ggplot(data.senior, aes(y=rate.senior)) +
geom_chau_boxplot(fill="purple",width=3) +
theme(legend.position = "none") +
scale_x_discrete(breaks = NULL) +
ylim(-0.057,0.077) +
theme(plot.margin = unit(c(0, 0, 0, 0), "inches")) +
labs(title="C.boxplot", subtitle="Senior civil servants", x="", y="")
print(C.boxplot.senior)
References
Hongmei Lin, Riquan Zhang and Tiejun Tong (2025). When Tukey meets Chauvenet: a new boxplot criterion for outlier detection. Journal of Computational and Graphical Statistics, accepted.
Contact
Should you have any questions, please feel free to contact Tiejun Tong via [email protected].