materncov.bands: Matern correlation or covariance function approximate...

In inlabru: Bayesian Latent Gaussian Modelling using INLA and Extensions

Matern correlation or covariance function approximate credible bands.

Description

Evaluate the covariance function for an inla.spde objectPlots the posterior distribution of the range, log(range), variance, or log(variance) parameter of a model's SPDE component. Can also plot Matern correlation or covariance function.

Usage

materncov.bands(
  manifold,
  dist,
  log.range,
  log.variance = NULL,
  alpha = 2,
  quantile = 0.95,
  n = 64,
  S1.L = NULL
)

Arguments

manifold	Either "R1", "S1", "R2", or "S2", from fm_manifold(mesh), or a full fm_mesh_2d or fm_mesh_1d object.
dist	A vector of distances at which to calculate the covariances/correlations
log.range	A scalar or a list (mean, sd), such as produced by inla.spde.result(...)$summary.log.range.nominal[[1]][c("mean","sd")]
log.variance	Either NULL, a scalar, or vector of the same type as for log.range. When NULL, the correlations are calculated instead of the covariances.
alpha	The SPDE operator order. Default 2.
quantile	The target credible probability. Default 0.95.
n	The number of parameter combinations to use for the approximation. Default 64.
S1.L	For manifold "S1", give the length of the cyclic interval

Details

Uses a Gaussian assumption for the internal model parameters, and finds a region in parameter space with approximately quantile probability.

Value

A list with estimated covariance or correlation (when log.variance is NULL) functions:

lower	An approximate lower bound for the quantile credible region
median	The function for for the approximate median parameters quantile
upper	An approximate upper bound for the quantile credible region

Author(s)

Finn Lindgren [e-mail]

inlabru documentation built on April 3, 2025, 10:54 p.m.