In K-Molloy/tevt: Threshold Methods for Extreme Value Theory

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(tevt)

Adding the data to the current workspace

data(danish)

Hill Plots

The standard averaged hill plot

gpd.plot.hill1(danish) 

Alternative Hill plot over the interval $[0,1]$ with confidence intervals

gpd.plot.hill2(danish, conf.int = TRUE) 

QQ-Estimator Plot

gpd.qq.est(danish)

Double Bootstrap for OSF (Optimal Sample Fraction)

An Implementation of the procedure proposed in Gomes et al. (2012) and Caeiro et al. (2016) for selecting the optimal sample fraction in tail index estimation.

gomes = gpd.gomes(danish)
out = gpd.fit(danish, threshold=gomes$threshold)
gpd.diag(out, safe=TRUE)

Double Bootstrap for OSF (Optimal Sample Fraction)

n Implementation of the procedure proposed in Danielsson et al. (2001) for selecting the optimal sample fraction in tail index estimation.

danielsson = gpd.danielsson(danish)
out = gpd.fit(danish, threshold=danielsson$threshold)
gpd.diag(out, safe=TRUE)

Single Bootstrap for OSF (Optimal Sample Fraction)

An Implementation of the procedure proposed in Hall (1990) for selecting the optimal sample fraction in tail index estimation

hall = gpd.hall(danish)
out = gpd.fit(danish, threshold=hall$threshold)
gpd.diag(out, safe=TRUE)

Automated Approach for Hill-Plot

An Implementation of the so called Eye-balling Technique proposed in Danielsson et al. (2016)

eye = gpd.eye(danish)
out = gpd.fit(danish, threshold=eye$threshold)
gpd.diag(out, safe=TRUE)

Fitting the Minimum Distance model

Minimising the distance between the empirical tail and a theoretical Pareto tail with respect to k. An Implementation of the procedure proposed in Danielsson et al. (2016) for selecting the optimal threshold in extreme value analysis.

md = gpd.md(danish)
out = gpd.fit(danish, threshold=md$threshold)
gpd.diag(out, safe=TRUE)

Old Vignette stuff

Fitting the model

out <- gpd.fit(danish, threshold=25)

Cheap and easy diagnostic plots At some point, this needs to override S3 plot function

gpd.plot(out, pick=1)
gpd.plot(out, pick=2)
gpd.plot(out, pick=3)
#gpd.plot(out, pick=4)

Better diagnostic plots, differs whether model is stationary.

gpd.diag(out, safe=TRUE)

But each of these can be accessed independently using their associated function names

gpd.qq2(out)

gpd.hist(out)

gpd.rlPlot(out)

K-Molloy/tevt documentation built on Dec. 18, 2021, 2:34 a.m.