tcG.fit: Fitting meta-Gaussian model(s)

In mbtgy/tcG: Fitting meta-Gaussian models

Fitting meta-Gaussian model(s)

Description

This function allows fitting 4 meta-Gaussian models. The data should only contain positive values and have a discrete component in zero.

Usage

tcG.fit(
  y,
  name,
  init,
  ym = 0,
  step = 0,
  plots = TRUE,
  bootstrap = TRUE,
  R = 500,
  filename = NULL,
  ncpus = parallel::detectCores(),
  ym.param = FALSE
)

Arguments

y	vector of data
name	name of the anamorphosis that will be fitted, a subset of c('gp', 'power', 'power-exp', 'quadratic-power')
init	initialisation parameters, a list named with the name argument
ym	minimal value that can be observed
step	discretization step
plots	boolean, should plots be drawn?
bootstrap	boolean, should boostrap replicates be fitted?
R	number of bootstrap replicates
filename	if you want to save the plots, a character for your file's name
ncpus	number of cores
ym.param	boolean, should ym be a parameter?

Details

A meta-Gaussian model is a variable Y defined as

Y = 0 * I_{X<0} + \psi(X) * I_{X\ge0}

with X~N(\mu,1) and where I is the indicator function, equals to 1 if the condition is true and 0 else.

The available fun are the following:

• "gp" for GP meta-Gaussian as in \insertCiteboutigny2021modelling;textualtcG: \psi(x) = ym+ \sigma x^(1/\alpha) exp(\xi x^2/2)

• "power" as in \insertCitebardossy1992space;textualtcG: \psi(x) = ym + \sigma x^(1/\alpha)

• "power-exp" as in \insertCiteallard2015disaggregating;textualtcG: \psi(x) = ym + \sigma1 (exp(\sigma2 * x^(1/\alpha))-1)

• "quadratic-power" as in \insertCiteallcroft2003latent;textualtcG: \psi(x) = ym + \sigma1 * x^(1/\alpha) + \sigma2 * x^(2/\alpha)

Value

A list of four elements containing the results of the fitted model(s). The first three elements are lists named by the fitted model(s) (argument name). par gives the fitted parameters. AIC gives the Akaike criterion. fit.boot gives the parameters fitted with the bootstrap replicates. Finally for.plots is a list that is used by the function res.plot.

References

\insertAllCited

Examples

## Not run: 
data(rain, package = "ismev")

# The histogram is not very revelant due to the fact \code{rain} are daily observations
res=tcG.fit(rain, name=c("power", "gp"),
            init=list("power"=c(0,2,2), "gp"=c(0,2,0.5,0.5)),
            ym=0.3, step=0.1, R=50)

## End(Not run)

mbtgy/tcG documentation built on Oct. 19, 2023, 3:10 p.m.