nieve-package {nieve}

Miscellaneous Utilities for Extreme Value Analysis

Description

The DESCRIPTION file:

Package:	nieve
Type:	Package
Title:	Miscellaneous Utilities for Extreme Value Analysis
Version:	0.1.3
Authors@R:	c(person(given = "Yves", family = "Deville", role = c("cre", "aut"), email = "[e-mail]", comment = c(ORCID = "0000-0002-1233-488X")))
Maintainer:	Yves Deville <[e-mail]>
Description:	Provides utility functions and objects for Extreme Value Analysis. These include probability functions with their exact derivatives w.r.t. the parameters that can be used for estimation and inference, even with censored observations. The transformations exchanging the two parameterizations of Peaks Over Threshold (POT) models: Poisson-GP and Point-Process are also provided with their derivatives.
License:	GPL (>= 2)
Suggests:	testthat, numDeriv, Renext, knitr, covr
Encoding:	UTF-8
URL:	https://github.com/yvesdeville/nieve/
BugReports:	https://github.com/yvesdeville/nieve/issues/
RoxygenNote:	7.2.3
VignetteBuilder:	knitr
Author:	Yves Deville [cre, aut] (<https://orcid.org/0000-0002-1233-488X>)

Index of help topics:

Exp1                    Density, Distribution Function, Quantile
                        Function and Random Generation for the
                        One-Parameter Exponential Distribution
GEV                     Density, Distribution Function, Quantile
                        Function and Random Generation for the
                        Generalized Extreme Value (GEV) Distribution
GPD2                    Density, Distribution Function, Quantile
                        Function and Random Generation for the
                        Two-Parameter Generalized Pareto Distribution
                        (GPD)
PP2poisGP               Transform Point-Process Parameters into
                        Poisson-GP Parameters
nieve-package           Miscellaneous Utilities for Extreme Value
                        Analysis
poisGP2PP               Transform Poisson-GP Parameters into
                        Point-Process Parameters

The nieve package provides utility functions for Extreme Value Analysis. It includes the probability functions for the two-parameter Generalized Pareto Distribution (GPD) and for the three-parameter Generalized Extreme Value (GEV) distribution. These functions are vectorized w.r.t. the parameters and optionally provide the exact derivatives w.r.t. the parameters: gradient and Hessian which can be used in optimization e.g., to maximize the log-likelihood. Since the gradient is available for the distribution function, the exact gradient of the log-likelihood function is available even when censored observations are used.

These functions should behave like the probability functions of the stats package: when a probability p = 0.0 or p = 1.0 is given, the quantile functions should return the lower and the upper end-point, be they finite or not. Also when evaluated at -Inf and Inf the probability functions should return 0.0 and 1.0.

The nieve package was partly funded by the French Institut de Radioprotection et Sûreté Nucléaire (IRSN) and some of the code formerly was part of R packages owned by the IRSN Bureau d'Expertise en Hydrogéologie et sur les Risques d'Inondation, météorologiques et Géotechniques (Behrig).