Skip to contents

Restricting to the largest fourth of the data, returns the number of exceedances that minimizes the Kolmogorov-Smirnov statistic, i.e., the maximum absolute difference between the estimated generalized Pareto and the empirical distribution of exceedances. Relative to the paper, different estimation methods are proposed.

Usage

thselect.pickands(xdat, thresh, method = c("mle", "lmom", "quartiles"))

Arguments

xdat

[numeric] vector of observations

thresh

[numeric] vector of candidate thresholds. If missing, defaults to order statistics from the 10th to a quarter of the sample size.

method

[string] estimation method, either the quartiles of Pickands (1975), maximum likelihood, probability weighted moments or L-moments

Value

a list with components

  • k0: number of exceedances

  • thresh0: selected threshold returned by the procedure

  • thresh: vector of candidate thresholds

  • dist; vector of Kolmogorov-Smirnoff distance

  • method; string for the estimation method

  • scale: estimated scale parameter at the chosen threshold

  • shape: estimated shape parameter at the chosen threshold

Note

The quartiles estimator of Pickands is robust, but very inefficient. It is provided for historical reasons.

References

James Pickands III (1975). Statistical inference using extreme order statistics, Annals of Statistics, 3(1) 119-131, doi:10.1214/aos/1176343003