GLmom (version 1.3.1)

optim.glme.gev11: Multi-start optimization for GEV11 model

Description

Internal function that performs multi-start optimization for the non-stationary GEV11 model. First solves L-moment equations via nleqslv (Broyden method), then optionally refines with penalized optimization via optim. Selects the best solution using goodness-of-fit or penalized negative log-likelihood.

Usage

optim.glme.gev11(
  xdat,
  ntry = 10,
  ftol = 1e-06,
  pretheta = NULL,
  model = "gev11",
  pen = "beta",
  mu = mu,
  std = std,
  p = p,
  c1 = c1,
  c2 = c2,
  choose = NULL
)

Value

A list containing:

para.lme

L-moment based estimates (5 parameters)

precis

Precision of the best solution

para.glme

GLME estimates (if pen != "no")

nllh.glme

Penalized negative log-likelihood (if pen != "no")

pen

Penalty type used

Arguments

xdat

Numeric vector of data.

ntry

Number of random starting points. Default is 10.

ftol

Tolerance for convergence. Default is 1e-6.

pretheta

Pre-estimated parameter vector (mu0, mu1, sigma0, sigma1, xi).

model

Model type: "gev11" (default), "gev10", or "gev20".

pen

Penalty type: "beta" (default), "norm", "ms", "park", "cannon", "cd", or "no".

mu

Normal penalty mean.

std

Normal penalty standard deviation.

p

Beta penalty shape parameter.

c1

Beta penalty scaling parameter.

c2

Beta penalty limit parameter.

choose

Selection method: "gof" (goodness-of-fit) or "nllh".