convassess: Convergence Assessment for Fitted Objects

In POT: Generalized Pareto Distribution and Peaks Over Threshold

Convergence Assessment for Fitted Objects

Description

convassess is a generic function used to assess the convergence of the estimation procedure to the global maximum. The function invokes particular methods which depend on the class of the first argument. This function uses several starting values to assess the sensitiveness of the fitted object with respect to starting values.

Usage

convassess(object, n = 50)
  
## S3 method for class 'uvpot'
convassess(object, n = 50)
## S3 method for class 'mcpot'
convassess(object, n = 50)
## S3 method for class 'bvpot'
convassess(object, n = 50)

Arguments

object	A fitted object. When using the POT package, an object of class 'uvpot', 'mcpot' or 'bvpot'. Generally, an object return by fitgpd, fitmcgpd or fitbvgpd.
n	The number of starting values to be tested.

Details

The starting values are defined using the unbiased probability weighted moments fitted on n bootstrap samples.

Value

Graphics: the considered starting values, the objective values derived from numerical optimizations and traceplots for all estimated parameters. In addition, it returns invisibly all these informations.

Author(s)

Mathieu Ribatet

See Also

fitgpd, fitmcgpd, fitbvgpd

Examples

set.seed(1)
##Univariate Case
x <- rgpd(30, 0, 1, 0.2)
fgpd1 <- fitgpd(x, 0, "med")
convassess(fgpd1)

##Bivariate Case
x <- rbvgpd(50, model = "log", alpha = 0.5, mar1 = c(0, 1, 0.2))
fgpd2 <- fitbvgpd(x, c(0.5,0.5))
convassess(fgpd2)

POT documentation built on April 14, 2022, 5:07 p.m.