Predict: Predict

In italo-goncalves/geomod3D: Implicit 3D Geological Modeling and Geostatistics

Description

Predictions at new spatial locations.

Usage

Predict(object, ...)

## S4 method for signature 'GP'
Predict(object, target, to = "value", output.var = T)

## S4 method for signature 'GP_geomod'
Predict(object, target, to = "value", output.ind = T,
  output.prob = T, use.unknown = T, Nsamp = 10000)

## S4 method for signature 'SPGP'
Predict(object, target, to = "value", output.var = T)

Arguments

target	The spatial3DDataFrame object to receive the prediction.
to	The name of the column in which to write the prediction. Will be overwritten if it exists or created if not.
output.var	Should the predictive variance be computed?
output.ind	Return indicators for boundary drawing?
output.prob	Return class probabilities?
use.unknown	Include the unknown class in output?
Nsamp	Number of samples used to estimate class probabilities.

Details

GP and SPGP objects return the predicted mean and variance for each location in target. The sparse GP returns two variances: var_full represents the total prediction uncertainty while var_cor is the amplitude of variation of the underlying latent function. The proportion between the two varies according to the distance from the pseudo_inputs, and is given in the quality column.

The GP_geomod object will calculate an indicator and its variance for each class at each location, which jointly form a multivariate normal distribution of the true indicators (or log-transformed compositional coordinates). The returned probabilities are actually the proportion of this probability mass over the region in which each indicator is dominant. This quantity is approximated by drawing Nsamp samples from the distribution and computing the number of times each indicator is higher than the others. The use.unknown = F option simply drops the probability of the unknown class and re-normalizes the rest in order for them to add to 1.

Value

A 3D spatial object of the same class as target containing the predictions.

italo-goncalves/geomod3D documentation built on May 24, 2019, 2:49 p.m.