shape.lgm: Update for latent Gaussian model for the shape parameter of a...

In lbelzile/mgp: Multivariate generalized Pareto models

Update for latent Gaussian model for the shape parameter of a generalized Pareto

Description

The shape has aGaussian prior with variance shape.tausq and precision (the inverse of the correlation matrix) given by shape.precis.

Usage

shape.lgm(scale, shape, ldat, shape.mu = NULL, shape.precis = NULL,
  shape.tausq = NULL, mmax, lbound = -0.5, ubound = 0.5,
  discount = 1, maxstep = 0.1)

Arguments

scale	vector of scale parameters for the generalized Pareto distribution
shape	vector of shape parameters for the generalized Pareto distribution
ldat	list with exceedances at each site
shape.mu	mean of the Gaussian process for the shape parameters
shape.precis	precision matrix of the Gaussian process corresponding to the inverse of the correlation matrix
shape.tausq	variance of the Gaussian process for the shape parameters
mmax	vector of maximum of each series in ldat
lbound	lower bound if parameter is truncated below
ubound	upper bound if parameter is truncated above
discount	numeric giving the discount factor for the Newton Default to 1.
maxstep	maximum step size for the MCMC (proposal will be at most maxstep units away from the current value).

Value

a vector of scale parameters

lbelzile/mgp documentation built on Aug. 5, 2023, 2:34 a.m.