rfield: Simulate data from a 2-D field with non-radial basis

In hillarykoch/nrbfields: Regression-based surface approximation using non-radial basis functions

Description

Currently, this function can a field simulate from a regular grid of basis functions with normal errors.

Usage

rfield(gridlen, coefs, shape, rate, scale, meanlog, sdlog,
        nbasis, type = c("gamma", "weibull", "lognormal"), rev = FALSE,
        copula = FALSE, copulaType = NULL, param = NULL,
        errtype = c("gaussian"), errvar = 1)

Arguments

gridlen	Length of regular grid.
coefs	Numeric vector of coefficients, including intercept term (if no intercept, set first entry to 0). Length of coefs must be equal to nbasis^2.
shape	Shape parameter of gamma of weibull distributions.
rate	Rate parameter of gamma distribution.
scale	Scale parameter of weibull distribution.
meanlog	Mean of the normal distribution on the log scale
sdlog	Standard deviation of the normal distribution on the log scale
nbasis	Number of basis functions to create in 1 dimension (this gets squared if making a field).
rev	Logical. Should the direction of the basis functions be reverse? In 1-D, this correspond to letting y=f(-x).
copula	Logical. Should 2-D bases be made up of dependent 1-D components?
copulaType	A character vector specifying which type of copula should be used to model the dependence when copula = TRUE. Options include "clayton", "frank", "gumbel", "amh", and "joe". These are Archimedean copulas implemented in the copula package.
param	Numeric value of the copula dependence parameter.
errtype	What sort of error to add to the basis? Currently, only gaussian errors are available.
errvar	Numeric specifying the variance of the error distribution.

Details

Currently no support for irregular grid simulation (aside by selecting some coefficients to be 0).

Value

Matrix corresponding a linear combination of basis functions.

Author(s)

[e-mail]

See Also

archmCopula

Examples

library(VGAM)
set.seed(1)

nbasis <- 40
coefs <- VGAM::rgpd(nbasis^2+1, location = 1, scale = 10, shape = 0.75)
coefs[sample(seq_along(coefs), round(length(coefs)*.3), replace = FALSE)] <- 0

b <- create.weibull.field(80, 1.5, 3, nbasis, rev = TRUE, reduce = TRUE)
pure <- rfield(80, coefs = coefs, shape = 1.5, scale = 3, nbasis = nbasis,
                type = "weibull", rev = TRUE, errvar = 0)
noisy <- rfield(80, coefs = coefs, shape = 1.5, scale = 3, nbasis = nbasis,
                type = "weibull", rev = TRUE, errvar = 1)

par(mfrow = c(1,3))
contour(b)
contour(pure)
contour(noisy)

coefs2 <- VGAM::rgpd(nbasis^2+1, location = 0, scale = 5, shape = 0.65)
b2 <- create.lognormal.field(100, .5, 3, nbasis, rev=TRUE,reduce=TRUE, copula = TRUE, copulaType = "frank", param = -4)
pure2 <- rfield(80, coefs = coefs2, meanlog = .5, sdlog = 3, nbasis = nbasis,
                type = "lognormal", rev = TRUE, copula = TRUE, copulaType = "frank", param = -4, errvar = 0)
noisy2 <- rfield(80, coefs = coefs2, meanlog = .5, sdlog = 3, nbasis = nbasis,
                type = "lognormal", rev = TRUE,  copula = TRUE, copulaType = "frank", param = -4, errvar = .5)

image(b2^.3, col=colorRampPalette(c("white","blue"))(256), useRaster=TRUE)
image(pure2^.3, col=colorRampPalette(c("white","blue"))(256), useRaster=TRUE)
image(noisy2^.3, col=colorRampPalette(c("white","blue"))(256), useRaster=TRUE)

hillarykoch/nrbfields documentation built on May 25, 2019, 8:17 a.m.