Code for Choosing the threshold in extreme value analysis

This online repository contains the code for reproducing the results in Belzile and Davison (2026).

The Padova_exceedances.RData contains the Padova data (daily cumulated rainfall amounts abov 0.02mm) for July-September during the period 1878 to 2016.

The folder applications contains the code to generate the figures and tables for the application to the Padova data.

• fig-padova-shape.R contains the code to generate Figure 1.
• fig-padova_thstab.R generates Figures 2-5
• tab-padova_parametric.R generates Table 1
• tab-padova-semiparametric.R generates Table 2
• fig-padova_eda.R uses the full Padova data to generate Figure 6 (Figure S1 of the Supplementary material)
• fig-penultimate_normal.R generates Figure 7 (Figure S2 of the SM).
• tab-penultimate_behaviour.R generates Tables 3 and 4 (Tables S1 and S2)
• fig-data-illustrations.R generates the results of Appendix D (Table 6, or Table S4 of the SM)
• fig-egpd-retlev.R generates figures on comparison of GPD vs EGPD models (not shown in the paper)

The folder simulations contains the code used for the simulation studies, indexed by 1 (parametric) or 2 (semiparametric).

• simulation*.R: code to generate a table of results
• script*.R: launches the simulations using R package simsalapar.
• postprocess*.R: takes the output from the cluster and format them into tables
• results*.R: generates all tables and figures in the supplementary material pertaining to the simulation studies.

All simulations were run on the LACED clusters at HEC Montreal, using R version 4.2.1 along with packages simsalapar (1.0-13), mev (2.2), tea (1.1), threshr (1.0.8), actuar (3.3-7) and longevity (1.3) as well as some utilities from the unpublished threshold package (2025.10.24).