mgp: Simulate a Multivariate Gaussian process
In ErickChacon/day2day: Functions for Day-to-Day Data Analysis

Description Usage Arguments Details Value Author(s) Examples

mgp Simulate a Multivariate spatial Gaussian process known as linear model of coregionalization Y(h) = AS(h), where S(h) is a vector of q independent Gaussian processes.

1 2	mgp(s1, s2, cov.model = NULL, variance = NULL, nugget = NULL, phi = NULL, kappa = NULL)

`s1`	First coordinate
`s2`	Second coordinate
`cov.model`	A character or function indicating the covariance function that Should be used to compute the variance-covariance matrix
`variance`	A qxq matrix of non-spatial covariance.
`nugget`	A qxq diagonal matrix of non-spatial noise.
`phi`	A q-length vector of decay parameters.
`kappa`	A q-length vector of kappa parameters if Matern spatial correlation function is used.

details.

A vector of the realization of the Gaussian Process

Erick A. Chacon-Montalvan

# Generate coordinates
N <- 100
s1 <- 2 * runif(N)
s2 <- 2 * runif(N)

# Covariance parameters
q <- 2
var <- sqrt(diag(c(4, 4)))
A <- matrix(c(1, - 0.8, 0, 0.6), nrow = 2)
variance <- var %*% tcrossprod(A) %*% var
nugget <- diag(0, q)
phi <- rep(1 / 0.08, q)

# Generate the multivariate Gaussian process
y <- mgp(s1, s2, "exponential", variance, nugget, phi)
y1 <- y[1:N]
y2 <- y[(N + 1):(2 * N)]

# Check correlation
cor(y1, y2)
plot(y1, y2)

# Visualize the spatial
plot(s1, s2, cex = y1, col = 2)
points(s1, s2, cex = y2, col = 3)