simField: Simulate A continous 2D spatial field using SPDE

In nmmarquez/PointPolygon: Evaluate Point and Polygon Data Simulatneously

Description

Calculates a continous underlying spatial field for a given shape and uses an additive model to combine with covariates to create a final observed field which is a linear combination of covariates their beta values with an optional link function applied.

Usage

simField(N = 60, sigmaE = 1, rangeE = 0.3, rho = 0.95,
  shape = NULL, nTimes = 1, beta0 = 0, betaList = list(),
  link = arm::invlogit, ...)

Arguments

N	Number of width pixels for the projection
sigmaE	spatial variance
rangeE	spatial range
rho	temporal autocorrelation for AR1 process
shape	polygon object to populate, if NULL unit square is used
nTimes	number of time observations
beta0	itercept term added to all points
betaList	list of 2 item lists with a type item which is either random, cluster, or spatial and a value item which is the beta coefficient.
link	link function to apply to the linear combination
...	other parameters to pass to mesh

Value

field object, contains 7 items. Spatial points data frame with raster values for transformed linear combination, and beta value. A mesh that was used to create the latent field and possibly covariates. The latent field itself. A bounding shape where all observations take place. A projection matrix from the mesh to the entire field of interest. The spde for the matern approximation. The beta coefficients used to produce the underlying field.

Examples

## Not run: 
require(tidyr)
require(gridExtra)
require(ar.matrix)
require(dplyr)
require(ggplot2)

unitSim <- simField(
    N = 500, 
    offset = c(0.1, 0.2), 
    max.edge = c(0.05,0.2),
    beta0 = -2,
    betaList = list(
        list(type="random", value=2),
        list(type="spatial", value=-.5),
        list(type="cluster", value=-2)
    ))

plotList <- lapply(c("V1", "V2", "V3", "theta"), function(eff){
    unitSim$spdf@data %>%
        dplyr::select(-V0, -z) %>%
        gather("Effect", "Value", V1:theta) %>%
        filter(Effect==eff) %>%
        ggplot(aes(x, y, fill=Value)) +
        geom_raster() +
        coord_equal() +
        theme_void() +
        scale_fill_distiller(palette = "Spectral") +
        ggtitle(eff)
})

do.call(grid.arrange, c(plotList, ncol=2))

usSim <- simField(
    N = 600,
    shape = US.df,
    rangeE = 1.7,
    offset = c(1, 2),
    max.edge = c(.5, 1))

plot(usSim$mesh)
usSim$spdf@data  %>%
    ggplot(aes(x, y, fill=z)) +
    geom_raster() +
    coord_equal() +
    theme_void() +
    scale_fill_distiller(palette = "Spectral")


## End(Not run)

nmmarquez/PointPolygon documentation built on Dec. 10, 2020, 1:15 a.m.