predict.ngme.spatial: Prediction.

In davidbolin/ngme: Linear Mixed Effects Models With Flexible Distributions

Prediction.

Description

Spatial prediction and cross-validation.

Usage

## S3 method for class 'ngme.spatial'
predict(
  object,
  id = NULL,
  type = "Smoothing",
  data = NULL,
  quantiles = c(0.025, 0.975),
  controls = list(return.samples = FALSE, excursions = NULL, crps = TRUE, nSim = 1000,
    nBurnin = 100, silent = TRUE, seed = NULL, n.cores = 1)
)

Arguments

object	A fitted object obtained by calling "ngme.spatial".
id	A numeric vector containing the ID's of the replicates for whom predictions are to be obtained. Default is set to "NULL" indicating perform predictions for all replicates.
type	A character string for the type of prediction: "Smoothing" gives spatial prediction based on all available data, "LOOCV" gives leave-one-out crossvalidation where also a number of accuracy measures are calculated. For multivariate models, "LOOCV" gives cross-validation over observation locations (so the data for all dimensions are removed and predicted), whereas "LOOCV" gives crossvalidation over univariate observations.
quantiles	A two-element vector that contains the quantiles of the predictions to be calculated.
controls	A list of control variables. "return.samples" A logical variable for returning the Monte Carlo samples used to compute the predictions; "TRUE" indicates return, "FALSE" do not return. "excursions" A list of excursion probabilities to compute. Each list should contain: "type" Type of excursion for indicating calculation of the probabilities, greater than or less than a threshol. User should specify "greater than" by ">" "less than" by "<". "level" A numeric value for the threshold compute excursion probability for. "process" A character string that for which component of the model the excursion probabilities to be calculated. 'X' for latent process plus fixed and random effects, 'W' for latent process, 'Y' for data, "crps" A logical variable for calculating continuous ranked probability score (CRPS); "TRUE" indicates calculate, "FALSE" do not calculate. "nSim" A numerical value for the number of samples for the Gibbs sampler to obtain the predictions. "nBurnin" A numeric value for the number of samples that are discarded as burn-in whilst calculating the predictions. silent A logical value for printing the details; "TRUE" indicates do not print, "FALSE" indicates print. n.cores Number of cores to use for paralell computations, default is set to 1. If larger than one, predictions over different replicates are computed in paralell.

Value

A list of output.

See Also

ngme.spatial

Examples

  ## Not run: 
  fit <- ngme.spatial(...)
  predict(fit, ...)
  
## End(Not run)

davidbolin/ngme documentation built on Dec. 5, 2023, 11:48 p.m.