In f-santi/tukeyGH: Tukey's g-and-h Probability Distribution

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "README-",
  eval = FALSE
)

Tukey's g-and-h Probability Distribution

Presentation

The package provides distribution, density and quantile functions of the Tukey's g-and-h probability distribution [@tukey1977], as well as functions for random number generation, parameter estimation and testing.

In the current version of the package, the g-and-h distribution can be fitted through:

indirect inference [@bee2019a]
the quantile estimator [@hoaglin1985]
maximum likelihood [@bee2019b]

whereas the hypothesis $h=0$ (which makes the g-and-h distribution, a g distribution) is tested by means of the log-likelihood ratio test procedure proposed in @bee2021b.

An example

Fit the g-and-h distribution to dataset EPWS2014 on operational losses by means of indirect inference [@bee2019a], and quantile estimator [@hoaglin1985]:

library(tukeyGH)
data("EPWS2014")

modII <- fitGH(EPWS2014, method = "iinference")
summary(modII)

modQU <- fitGH(EPWS2014, method = "quantile")
summary(modQU)

modML <- fitGH(EPWS2014, method = "mle")
summary(modML)

rbind(QU = coef(modQU), II = coef(modII), ML = coef(modML))