SDALGCP (version 0.4.0)

SDALGCPPred: Spatial prediction using plug-in of MCML estimates

Description

This function performs spatial continuous and discrete prediction, fixing the model parameters at the Monte Carlo maximum likelihood estimates of a SDALGCP model.

Usage

SDALGCPPred(
  para_est,
  cellsize,
  continuous = TRUE,
  control.mcmc = NULL,
  pred.loc = NULL,
  divisor = 1,
  plot.correlogram = F,
  messages = TRUE,
  parallel = FALSE
)

Arguments

para_est

an object of class "SDALGCP" obtained as a result of a call to SDALGCPMCML.

cellsize

the size of the computational grid

continuous

logical; to choose which prediction to do perform, discrete or continuous. the default is continuous.

control.mcmc

output from controlmcmcSDA, if not provided, it uses the values used for the parameter estimation

pred.loc

optional, the dataframe of the predictive grid.

divisor

optional, the value to use to convert the dimension of the polygon, default is 1 which implies no conversion

plot.correlogram

logical; if plot.correlogram=TRUE the autocorrelation plot of the conditional simulations is displayed.

messages

logical; if messages=TRUE then status messages are printed on the screen (or output device) while the function is running. Default is messages=TRUE.

parallel

to parallelize some part of the function.

Value

pred.draw: the samples of the prediction

pred: the prediction of the relative risk

predSD: the standard error of the prediction

Pred.loc: The coordinates of the predictive locations

Details

The function perform prediction of the spatially discrete incidence and covariate adjusted relative risk, and spatially continuous relative risk. The discrete inference uses the Metropolis-Adjusted Langevin Hasting sampling from Laplace.sampling. And the continuous inference is typically change of support inference.

References

Banerjee, S., Carlin, B. P., & Gelfand, A. E. (2014). Hierarchical modeling and analysis for spatial data. CRC press.

See Also

plot.Pred.SDALGCP, SDAContinuousPred, SDADiscretePred, plot_continuous, plot_discrete

Examples

# NOT RUN {
### Prepare the input of the model
data(PBCshp)
data <- as.data.frame(PBCshp@data)  #get the data
### Write the formula of the model
FORM <- X ~ propmale + Income + Employment + Education + Barriers + Crime +
Environment +  offset(log(pop))
### set the discretised phi
phi <- seq(500, 1700, length.out = 20)
#### get the initial parameter
model <- glm(formula=FORM, family="poisson", data=data)
beta.start <-coef(model)
sigma2.start <- mean(model$residuals^2)
phi.start <- median(phi)
par0 <- c(beta.start, sigma2.start, phi.start)
# setup the control arguments for the MCMC
n <- 545
h <- 1.65/(n^(1/6))
control.mcmc <- controlmcmcSDA(n.sim = 10000, burnin = 2000,
                 thin= 8, h=h, c1.h = 0.01, c2.h = 1e-04)
###Run the model
# }
# NOT RUN {
my_est <- SDALGCPMCML(formula=FORM, data=data, my_shp=PBCshp, delta=100, phi=phi, method=1,
                     weighted=FALSE,  plot=TRUE, par0=par0, control.mcmc=control.mcmc)
Con_pred <- SDALGCPPred(para_est=my_est,  cellsize=300, continuous=TRUE)
# }