GEVcdn-package: GEV Conditional Density Estimation Network

Description Details Author(s) References See Also


Parameters in a Generalized Extreme Value (GEV) distribution are specified as a function of covariates using a conditional density estimation network (CDN), which is a probabilistic variant of the multilayer perceptron neural network. If the covariate is time or is dependent on time, then the GEV CDN model can be used to perform nonlinear, nonstationary GEV analysis of hydrological or climatological time series. Owing to the flexibility of the neural network architecture, the model is capable of representing a wide range of nonstationary relationships, including those involving interactions between covariates. Model parameters are estimated by generalized maximum likelihood, an approach that is tailored to the estimation of GEV parameters from geophysical time series.


Procedures for fitting GEV CDN models are provided by the functions and gevcdn.bag. Once a model has been developed, gevcdn.evaluate is used to evaluate the GEV distribution parameters as a function of covariates. Confidence intervals for GEV parameters and specified quantiles can be estimated using gevcdn.bootstrap. All other functions are used internally and should not normally need to be called directly by the user.

Note: the GEV distribution functions are provided by the VGAM package. The convention for the sign of the shape parameter is opposite to that used in hydrology and thus differs from Cannon (2010).


Alex J. Cannon

Maintainer: Alex J. Cannon <[email protected]>


Cannon, A.J., 2010. A flexible nonlinear modelling framework for nonstationary generalized extreme value analysis in hydroclimatology. Hydrological Processes, 24: 673-685. DOI: 10.1002/hyp.7506

Cannon, A.J., 2011. GEVcdn: an R package for nonstationary extreme value analysis by generalized extreme value conditional density estimation network, Computers & Geosciences, 37: 1532-1533. DOI: 10.1016/j.cageo.2011.03.005

See Also


GEVcdn documentation built on May 29, 2017, 9:25 a.m.