Package index • ExtremeCI

Profile CIs using peaks over threshold (generalized Pareto distribution)

In most cases, if the data is not aggregated as block-maxima, the generalized Pareto distribution is the best choice for extrapolating quantile estimates beyond the range of observations as it makes better use of available data, at the cost of needing to select an appropriate threshold value.

GPD_profile_loglik_curve(): GPD profile log-likelihood curve
GPD_profile_CI(): GPD profile CI using binary search
GPD_profile_CIs_multiple(): Multi-value conditional GPD profile likelihood confidence intervals
GPD_profile_loglik(): GPD profile log-likelihood
GPD_maxlik(): Maximum-likelihood GPD estimate
GPD_quantiles(): Compute extreme quantile from GPD parameters
GPD_endpoint(): GPD endpoint
GPD_log_likelihood(): GPD log likelihood
GPD_change_parametrization(): GPD parameter vector reparametrization

Profile CIs using block maxima (generalized extreme value distribution)

When the data at hand is aggregated as block maxima (e.g. yearly maxima), the generalized extreme value distribution is the natural choice to extrapolate quantile estimates beyond the range of observations.

GEV_profile_loglik_curve(): GEV profile log-likelihood curve
GEV_profile_CI(): GEV profile CI using binary search
GEV_profile_CIs_multiple(): Multi-value conditional GEV profile likelihood confidence intervals
GEV_profile_loglik(): GEV profile log-likelihood
GEV_maxlik(): Maximum-likelihood GEV estimate
GEV_return_level(): Compute return level from GEV parameters
GEV_endpoint(): GEV endpoint
GEV_log_likelihood(): GEV log likelihood
GEV_change_parametrization(): GEV parameter vector reparametrization

Plotting helpers

Helper functions for plotting the profile log-likelihood curves and nonstationary quantile estimates with confidence intervals, as a function of data, using ggplot2.

plot_profile_loglik_curve(): Profile likelihood curve plot
plot_data_quantile_ci(): Non-stationary confidence bands plot