theta_hier: Extremal index estimation

In mickwar/mwEVT:

Description

Fill

Usage

theta_hier(y, u, n, prior, likelihood, K = 1, chain_init, ...)

Arguments

y	Numeric, sequence of depedent random variables. May be a matrix. See details.
u	Numeric, the threshold. Defaults to quantile(y, 0.90).
prior	List.
likelihood	Character, either "ferro" or "suveges" (default).
K	Numeric, the run parameter for likelihood = "suveges" only. Defaults to K = 1, the model proposed in Suveges (2007).
...	Additional arguments passed to mwBASE::mcmc_sampler.
ord	Numeric, vector of length R = NCOL(y). Defaults to 1:R. See details.

Details

The data y is expected to be observed values along an evenly-spaced grid of width 1. If y is an n x R matrix, then it will be collapsed to y = c(y[,ord]). This may be the case when working with multiple realizations from the same process or with specific seasons on a time-series.

The 'classical' option for method produces the estimates given by either Ferro and Segers (2003) or Suveges and Davison (2010). When method == 'bayesian' the likelihoods given in the aforementioned papers are used to produce posterior samples for theta.

mickwar/mwEVT documentation built on May 22, 2019, 9:56 p.m.