gp_fit {nimblewomble}R Documentation

Fit a Gaussian process

Description

Fits a Gaussian process with the choice of three kernels. Uses 'nimble' to generate posterior samples.

Usage

gp_fit(
  coords = NULL,
  y = NULL,
  X = NULL,
  kernel = c("matern1", "matern2", "gaussian"),
  niter = NULL,
  nburn = NULL
)

Arguments

coords

spatial coordinats (supply as a matrix)

y

response

X

covariates (supply as a matrix without the intercept)

kernel

choice of kernel; must be one of "matern1", "matern2", "gaussian"

niter

number of iterations

nburn

burn-in

Value

A list of MCMC samples containing the covariance parameters and the parameter estimates with associated 95

Author(s)

Aritra Halder <aritra.halder@drexel.edu>,
Sudipto Banerjee <sudipto@ucla.edu>

Examples


require(nimble)
require(nimblewomble)

set.seed(1)
# Generated Simulated Data
N = 1e2
tau = 1
coords = matrix(runif(2 * N, -10, 10), ncol = 2)
colnames(coords) = c("x", "y")
y = rnorm(N, mean = 20 * sin(sqrt(coords[, 1]^2  + coords[, 2]^2)), sd = tau)
# Posterior samples for theta
mc_sp = gp_fit(coords = coords, y = y, kernel = "matern2")
mc_sp$estimates


[Package nimblewomble version 0.1.0 Index]