spatial.pred.lm.ps {PrevMap}

Spatial predictions for the geostatistical Linear Gaussian model using plug-in of ML estimates

Description

This function performs spatial prediction, fixing the model parameters at the maximum likelihood estimates of a linear geostatistical model.

Usage

spatial.pred.lm.ps(
  object,
  grid.pred = NULL,
  predictors = NULL,
  predictors.intensity = NULL,
  control.mcmc = NULL,
  target = 3,
  type = "marginal",
  quantiles = NULL,
  standard.errors = FALSE,
  messages = TRUE,
  return.samples = FALSE
)

Arguments

object

an object of class "PrevMap" obtained as result of a call to linear.model.MLE.

grid.pred

a matrix of prediction locations. Default is grid.pred=NULL, in which case the grid used to approximate the intractable integral in the log-Gaussian Cox process model is used for prediction.

predictors

a data frame of the values of the explanatory variables at each of the locations in grid.pred, for the response variable model; each column correspond to a variable and each row to a location. Warning: the names of the columns in the data frame must match those in the data used to fit the model. Default is predictors=NULL for models with only an intercept.

predictors.intensity

a data frame of the values of the explanatory variables at each of the locations in grid.pred, for the log-Gaussian Cox process model; each column correspond to a variable and each row to a location. Warning: the names of the columns in the data frame must match those in the data used to fit the model. Default is predictors=NULL for models with only an intercept.

control.mcmc

output from control.mcmc.MCML which defined the control parameters of the Monte Carlo Markv chain algorithm.

target

an integeter indicating the predictive target: target=1 if the predictive target is the linear predictor of the response; target=2 is the predictive target is the sampling intensity of the preferentially sampled data; target=3 if both of the above are the predictive targets. Default is target=3.

type

a character indicating the type of spatial predictions for target=1: type="marginal" for marginal predictions or type="joint" for joint predictions. Default is type="marginal". Note that predictions for the sampling intensity (target=2) are always joint.

quantiles

a vector of quantiles used to summarise the spatial predictions.

standard.errors

logical; if standard.errors=TRUE, then standard errors for each scale.predictions are returned. Default is standard.errors=FALSE.

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.

return.samples

logical; if return.samples=TRUE a matrix of the predictive samples for the prediction target (as specified in target) are returned in the output.

Value

A "pred.PrevMap.ps" object list with the following components: response (if target=1 or target=3) and intensity (if target=2 pr target=3). grid.pred prediction locations. Each of the components intensity and response is a list with the following components:

predictions: a vector of the predictive mean for the corresponding target.

standard.errors: a vector of prediction standard errors (if standard.errors=TRUE).

quantiles: a matrix of quantiles of the resulting predictions with each column corresponding to a quantile specified through the argument quantiles.

samples: a matrix corresponding to the predictive samples of the predictive target (only if return.samples=TRUE), with each row corresponding to a samples and column to a prediction location. In the case of a model with a mix of preferential and non-preferential data, if target=1 or target=3, each of the above components will be a list with two components, namely preferential and non.preferential, associated with response.

Author(s)

Emanuele Giorgi [e-mail]