NSGEV-package.RdUtility functions for Non-Stationary time series with GEV margins.
The DESCRIPTION file:
This package was not yet installed at build time.
This package is dedicated to some Non-Stationary Extreme Value models with Generalised Extreme Value (GEV) marginal distributions. Such models are often used to investigate climate change from series of annual maxima.
An emphasis is put on Profile Likelihood inference on the parameters, quantiles or return levels. For that aim constrained optimisations are used, relying on the nloptr package.
The package provides most of the classical S3 methods useful to cope
with fitted model objects, such as predict, coef,
vcov, confint, residuals, logLik, ...
"NSGEV" This class was intended to be the main class
when the package was created. As a major difference with other EV
packages, the model parameters can be used in several of the GEV
parameters loc, scale and shape. A typical
example is when a scale proportional to the location needs to be
specified.
"TVGEV" This class is devoted to the special case of
Time Varying (TV) models using only functions of the time as
covariates. Specialised methods make sense to plot, predict and
more.
Most efforts have been put on the "TVGEV" class. In the future,
the "NSGEV" class could be discarded.
Coles, S. (2001) Introdution to Statistical Modelling of Extreme Values, Springer.
library(NSGEV)
methods(class = "TVGEV")
#> [1] anova autoplot bs cdf cdfMaxFun
#> [6] coef confint density logLik mean
#> [11] modelMatrices moment plot predict print
#> [16] profLik psi2theta quantMax quantMaxFun quantile
#> [21] residuals simulate summary vcov
#> see '?methods' for accessing help and source code