Ren2gumbel: Translate a vector of coefficients from a Renewal-POT model with exponential excesses to a vector of Gumbel parameters

Description

Translate a vector of coefficients from a Renewal-POT model with exponential excesses to a vector of Gumbel parameters.

Usage

Ren2gumbel(object,
           threshold = NULL,
           w = 1,
           distname.y = c("exponential", "exp"),
           jacobian = (length(w) == 1L),
           vcovRen = NULL)

Value

A vector of GEV parameters if w has length 1, and a matrix if

w has length > 1. The returned objects has attributes.

Arguments

object

A named vector of parameters or an object of class "Renouv". In the first case, the names of the vector element must conform to the exponential distribution so the vector must be of length 2 with names "lambda" and "rate".

threshold

A threshold associated with the parameters. If object is an object with class "Renouv", its threshold slot will be used.

w

A block duration or a vector of block durations.

distname.y

The name of the distribution for the excesses. Can be either "exponential" or "exp". The choice has no impact on the computations, but this name will be attached to the result as an attribute and may affect later use.

jacobian

Logical. If TRUE the jacobian matrix of the transformation will be computed and attached to the result as an attribute.

vcovRen

A covariance matrix for the Renouv parameters.

Author

Yves Deville

Examples

Run this code

## Fit a Renouv model with exponential excesses (default)
fit <- Renouv(Garonne)
## Convert to gumbel (usable for one-year block maxima)
parGumbel <- Ren2gumbel(fit)
## Retrieve the 'Renouv' model by giving the right threshold
parRen <- gumbel2Ren(parGumbel,
                     threshold = 2500,
                     vcovGumbel = attr(parGumbel, "vcov"),
                     plot = TRUE)
## Build a compatible model under the assumption of one event by
## year
parRen2 <- gumbel2Ren(parGumbel,
                      lambda = 1.00,
                      vcovGumbel = attr(parGumbel, "vcov"),
                      plot = TRUE)
parRenNames <- c("lambda", "rate")
## Build a 'Renouv' object without estimation
myVcov <- attr(parRen, "vcov")[parRenNames, parRenNames]
fitNew <- RenouvNoEst(threshold = attr(parRen, "threshold"),
                      estimate = parRen,
                      distname.y = "exp",
                      cov = myVcov)
## Compare return levels
cbind(roundPred(fit$pred)[ , -2], roundPred(fitNew$pred)[ , -2])
## idem for the putative 'Renouv' with rate 1
myVcov2 <- attr(parRen2, "vcov")[parRenNames, parRenNames]
fitNew2 <- RenouvNoEst(threshold = attr(parRen2, "threshold"),
                       estimate = parRen2,
                       distname.y = "exp",
                       cov = myVcov2)
cbind(roundPred(fit$pred)[ , -2], roundPred(fitNew2$pred)[ , -2])