predict.sbss: Predict Method for an Object of Class 'sbss'

Description

predict.sbss predicts the estimated source random field on a grid with Inverse Distance Weighting (IDW) and plots these predictions.

Usage

# S3 method for sbss
predict(object, p = 2, n_grid = 50, which = 1:ncol(object$s), ...)

Value

The return is dependent on the class of the latent field in the 'sbss' object. If class(object$s) is a matrix then a list with the following entries is returned:

vals_pred_idw: a matrix of dimension c(n,p) (when which is default or less than p columns according to the selected components with the which argument) with the IDW predictions of the estimated source random field.
coords_pred_idw: a matrix of dimension c(n,2) with the grid coordinates for the IDW predictions.

If class(object$s) is SpatialPointsDataFrame or sf then the predicted values and their coordinates are returned as an object of the corresponding class.

The return is invisible.

Arguments

object: object of class 'sbss'. Usually result of sbss.
p: numeric. The positive power parameter for IDW. Default is 2.
n_grid: numeric. Each dimension of the spatial domain is divided by this integer to derive a grid for IDW predictions. Default is 50.
which: a numeric vector indicating which components of the latent field should be predicted.
...: further arguments to the plot method of class(x$s), which is either spplot or plot.

Details

IDW predictions are made on a grid. The side lengths of the rectangular shaped grid cells are derived by the differences of the rounded maximum and minimum values divided by the n_grid argument for each column of object$coords. Hence, the grid contains a total of n_grid ^ 2 points. The power parameter of the IDW predictions is given by p (default: 2).

The predictions are plotted with the corresponding plot method of class(x$s). Either spplot for class(x$s) is SpatialPointsDataFrame or plot.sf for class(x$s) is sf. If x$s is a matrix then it is internally cast to SpatialPointsDataFrame and spplot is used for plotting. Arguments to the corresponding plot functions can be given through ... as it is done by the method plot.sbss.

Examples

Run this code

# simulate coordinates
coords <- runif(1000 * 2) * 20
dim(coords) <- c(1000, 2)
coords_df <- as.data.frame(coords)
names(coords_df) <- c("x", "y")
# simulate random field
if (!requireNamespace('gstat', quietly = TRUE)) {
  message('Please install the package gstat to run the example code.')
} else {
  library(gstat)
  model_1 <- gstat(formula = z ~ 1, locations = ~ x + y, dummy = TRUE, beta = 0, 
                   model = vgm(psill = 0.025, range = 1, model = 'Exp'), nmax = 20)
  model_2 <- gstat(formula = z ~ 1, locations = ~ x + y, dummy = TRUE, beta = 0, 
                   model = vgm(psill = 0.025, range = 1, kappa = 2, model = 'Mat'), 
                   nmax = 20)
  model_3 <- gstat(formula = z ~ 1, locations = ~ x + y, dummy = TRUE, beta = 0, 
                   model = vgm(psill = 0.025, range = 1, model = 'Gau'), nmax = 20)
  field_1 <- predict(model_1, newdata = coords_df, nsim = 1)$sim1
  field_2 <- predict(model_2, newdata = coords_df, nsim = 1)$sim1
  field_3 <- predict(model_3, newdata = coords_df, nsim = 1)$sim1
  field <- as.matrix(cbind(field_1, field_2, field_3))

  # apply sbss with three ring kernels
  kernel_borders <- c(0, 1, 1, 2, 2, 4)
  res_sbss <-  sbss(field, coords, 'ring', kernel_borders)
  
  # predict latent fields on grid with default settings
  predict(res_sbss)
  
  # predict latent fields on grid with custom plotting settings
  predict(res_sbss, colorkey = TRUE, as.table = TRUE, cex = 1)
  
  # predict latent fields on a 60x60 grid
  predict(res_sbss, n_grid = 60, colorkey = TRUE, as.table = TRUE, cex = 1)
  
  # predict latent fields with a higher IDW power parameter
  predict(res_sbss, p = 10, colorkey = TRUE, as.table = TRUE, cex = 1)
  
  # predict latent fields and save the predictions
  predict_list <- predict(res_sbss, p = 5, colorkey = TRUE, as.table = TRUE, cex = 1)
}