Skip to content

dtr0x/bias-corrected-pot-estimation-cvar

Repository files navigation

Bias-Corrected Peaks-Over-Threshold Estimation of the CVaR (UAI 2021)

https://proceedings.mlr.press/v161/troop21a.html

This code will generate all data used in simulations presented in the paper and produce plots. Parameters can be adjusted to produce results for other scenarios.

Requirements:

  • Python v3.xx
  • Modules:
    • numpy
    • scipy
    • matplotlib
    • multiprocessing

Replicating results

Executing run_sim.py will generate all samples from all distributions given in the paper, and compute estimated CVaRs (alpha=0.998) using UPOT, BPOT, and SA. All samples and CVaR values will be stored in the data folder. The following files can be used to reproduce plots in the paper:

  • dist_plots.py Produces RMSE and absolute bias plots from the generated CVaR data
  • coverage_prob.py Produces plots of coverage probability from the generated CVaR data
  • asymp_var.py Comparison of UPOT and SA asymptotic variance for the Frechet distribution

About

Bias-Corrected Peaks-Over-Threshold Estimation of the CVaR.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages