make_cov_fct: Generate covariance matrix function from covariance function

In larslau/pavpop: Phase and amplitude varying population pattern models

Description

Function that takes a covariance function and returns a function that generates covariance matrices according to the given covariance function.

Usage

make_cov_fct(cov_fct, noise = TRUE, param = NULL, ns = NULL, ...)

Arguments

cov_fct	covariance function.
noise	logical. Should an identity matrix be added to the covariance matrix?
param	standard values of parameters
ns	list of arguments to make a stationary covariance locally adaptive. The list has takes the following entries knots (mandatory) the number of parameters to control the non-stationary, fixed (optional) vector of fixed knots with reference weight of 1, if not supplied the last knot is chosen. See examples and Details.
...	arguments passed to cov_fct.

Examples

# Generate observation points
t <- seq(0, 1, length = 3)

# Generate covariances
matern_cov <- make_cov_fct(Matern, noise = FALSE)
bm_cov <- make_cov_fct(Brownian, noise = FALSE)
bb_cov <- make_cov_fct(Brownian, noise = FALSE, type = 'bridge')

# Evaluate covariance matrices
matern_cov(t, param = c(1, 1, 1))
bm_cov(t, param = c(1, 1, 1))
bb_cov(t, param = c(1, 1, 1))

# Plot covariance matrices
t <- seq(0, 1, length = 30)

persp(t, t, matern_cov(t, param = c(1, 0.5, 1)),
      theta = -30, phi = 30, ticktype = 'd', zlab = '',
      col = 'lightblue', shade = 0.2,
      main = 'Matern covariance (1, 0.5, 1)', zlim = c(0, 1))

persp(t, t, matern_cov(t, param = c(1, 0.5, 2)),
      theta = -30, phi = 30, ticktype = 'd', zlab = '',
      col = 'lightblue', shade = 0.2,
      main = 'Matern covariance (1, 0.5, 2)', zlim = c(0, 1))

persp(t, t, matern_cov(t, param = c(1, 0.1, 2)),
      theta = -30, phi = 30, ticktype = 'd', zlab = '',
      col = 'lightblue', shade = 0.2,
      main = 'Matern covariance (1, 0.1, 2)', zlim = c(0, 1))

persp(t, t, bm_cov(t, param = 1), theta = -30, phi = 30,
ticktype = 'd', zlab = '', col = 'lightblue', shade = 0.2,
main = 'Brownian motion covariance (tau = 1)', zlim = c(0, 1))

persp(t, t, bb_cov(t, param = 2), theta = -30, phi = 30,
ticktype = 'd', zlab = '', col = 'lightblue', shade = 0.2,
main = 'Brownian bridge covariance (tau = 2)', zlim = c(0, 1))

# Make covariance non-stationary
matern_cov_ns <- make_cov_fct(Matern, noise = FALSE,
                              ns = list(knots = 3, range = c(0, 1)))
# Use mid-point reference instead of last
matern_cov_ns_mid <- make_cov_fct(Matern, noise = FALSE,
                              ns = list(knots = 3, fixed = 2,
                                        range = c(0, 1)))

# Original covariance
persp(t, t, matern_cov_ns(t, param = c(1, 1, 1, 0.3, 2)),
      theta = -30, phi = 30, ticktype = 'd', zlab = '',
      col = 'lightblue', shade = 0.2,
      main = 'Matern covariance (1, 0.1, 2)', zlim = c(0, 1))

# Modified covariances
persp(t, t, matern_cov_ns(t, param = c(0.5, 0.7, 1, 0.3, 2)),
      theta = -30, phi = 30, ticktype = 'd', zlab = '',
      col = 'lightblue', shade = 0.2,
      main = 'Non-stationary Matern covariance (1, 0.1, 2)', zlim = c(0, 1.1))

persp(t, t, matern_cov_ns(t, param = c(1, 0.7, 1, 0.3, 2)),
      theta = -30, phi = 30, ticktype = 'd', zlab = '',
      col = 'lightblue', shade = 0.2,
      main = 'Non-stationary Matern covariance (1, 0.1, 2)', zlim = c(0, 1.1))

persp(t, t, matern_cov_ns_mid(t, param = c(1, 0.7, 1, 0.3, 2)),
      theta = -30, phi = 30, ticktype = 'd', zlab = '',
      col = 'lightblue', shade = 0.2,
      main = 'Non-stationary Matern covariance (1, 0.1, 2)', zlim = c(0, 1.1))

larslau/pavpop documentation built on June 14, 2019, 2:18 p.m.