gpd.prof: Profile likelihood based confidence intervals for GPD

gpd.profR Documentation

Profile likelihood based confidence intervals for GPD

Description

Calculates profile likelilhood based confidence intervals for a given fitted GPD model – this is only implemented for two parameter GPD with no covariates in the model.

Usage

gpd.prof(
  z,
  m,
  xmax,
  xlow,
  conf = 0.95,
  nint = 50,
  PlotIt = FALSE,
  mult = 2,
  priorParameters = NULL
)

Arguments

z

a fitted evmOpt object

m

return period : units are number of observations

xmax

point estimate of the return level, this is used to bracket the roots of the equation used to calculate the ends of the profile likelihood based confidence interval. The value need not be exact.

xlow

value lower than the lower end of the confidence interval, for bracketing in root finding

conf

confidence level, defaults to 0.95

nint

used for plotting if required, number of points at which to calculate the profile likelihood for plotting, defaults to 50

PlotIt

logical, whether or not to plot the profile likelihood, defaults to FALSE

mult

used to calculate the starting point for the root finding for solving to find the upper end of the confidence interval. The starting point is mult*xmax minus the lower end point. If this starting point is beyond the estimated upper endpoint of the fitted distribution then this can cause an error, and the value of mult should be reduced

priorParameters

optional, value of prior/penalty parameters used for penalised likelihood estimation, default to NULL

Value

Numeric vector of length two, with lower and upper ends of the estiamted confidence intervals respectively.


texmex documentation built on June 22, 2024, 12:26 p.m.