Goals and scope

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 IRSN/Behrig.

The nieve package is intended to be a “low-level” package, providing fast and well-tested “basic” functions for Extreme Value Analysis (EVA). It is not intended to provide sophisticated EVA models which should be found or be implemented in other packages.

The package

• Provides the derivatives w.r.t. the parameters for the probability functions related to the Generalized Pareto (GP) and the Generalized Extreme Value (GEV) distributions. This involves: the log-density, the distribution function or survival and the quantile function. The 2-nd order derivative (Hessian) is available in most cases.

• Provides probability functions which are vectorized w.r.t. the parameters, as required in non-stationary EV models or in Bayesian inference.

• Provides the transformations for the two usual parameterizations of Peaks Over Threshold (POT) models: Poisson-GP and Point Process (PP). The two transformations come with their derivatives, as needed to compute the covariance matrices.

Although several R packages devoted EVA compute the exact derivatives w.r.t. the parameters (extRemes, mev, …), to our best knowledge, none of these packages make the derivatives available via exported functions. See the vignette nieve: Yet Another Extreme Value Package? for more details.

The vignette Special Values in Extreme-Value Distributions compares the treatment of non-finite values in the probability functions across Extreme-Value packages.

Note that nieve has no special relation with the famous package snow:)

News

See the NEWS.md file or (after install) use news(package = "nieve").

Installation

From CRAN

nieve is on CRAN since 2022-12-02 hence can be installed using the install.packages function or the RStudio IDE.

From GitHub

Note that this package needs a compilation since it embeds some code written in C. If you are using Windows, you need to have the Rtools installed.

Provided that the devtools package is installed you can then in an R session use

devtools::install_github("yvesdeville/nieve", dependencies = TRUE)

You can also select a specific branch or a specific commit by using the suitable syntax for install_github, see the devtools package documentation.

Alternatively you can Clone the repository using git clone, and then build and install.