predict.sclm {RcppCensSpatial}

Prediction in spatial models with censored/missing responses

Description

It performs spatial prediction in a set of new S spatial locations.

Usage

## S3 method for class 'sclm'
predict(object, locPre, xPre, ...)

Arguments

object

object of class 'sclm' given as output of EM.sclm, MCEM.sclm, or SAEM.sclm function.

locPre

matrix of coordinates for which prediction is performed.

xPre

matrix of covariates for which prediction is performed.

...

further arguments passed to or from other methods.

Details

This function predicts using the mean squared error (MSE) criterion, which takes the conditional expectation E(Y|X) as the best linear predictor.

Value

The function returns a list with:

coord

matrix of coordinates.

predValues

predicted values.

sdPred

predicted standard deviations.

Author(s)

Katherine L. Valeriano, Alejandro Ordoñez, Christian E. Galarza, and Larissa A. Matos.

See Also

EM.sclm, MCEM.sclm, SAEM.sclm

Examples

set.seed(1000)
n = 120
coords = round(matrix(runif(2*n,0,15),n,2), 5)
x = cbind(rbinom(n,1,0.50), rnorm(n), rnorm(n))
data = rCensSp(c(1,4,-1), 2, 3, 0.50, x, coords, "left", 0.10, 20)

## Estimation
data1 = data$Data

# Estimation: EM algorithm
fit1 = EM.sclm(y=data1$y, x=data1$x, ci=data1$ci, lcl=data1$lcl,
               ucl=data1$ucl, coords=data1$coords, phi0=2.50, nugget0=1)

# Estimation: SAEM algorithm
fit2 = SAEM.sclm(y=data1$y, x=data1$x, ci=data1$ci, lcl=data1$lcl,
                 ucl=data1$ucl, coords=data1$coords, phi0=2.50, nugget0=1)

# Estimation: MCEM algorithm
fit3 = MCEM.sclm(y=data1$y, x=data1$x, ci=data1$ci, lcl=data1$lcl,
                 ucl=data1$ucl, coords=data1$coords, phi0=2.50, nugget0=1,
                 MaxIter=300)
cbind(fit1$theta, fit2$theta, fit3$theta)

# Prediction
data2 = data$TestData
pred1 = predict(fit1, data2$coords, data2$x)
pred2 = predict(fit2, data2$coords, data2$x)
pred3 = predict(fit3, data2$coords, data2$x)

# Cross-validation
mean((data2$y - pred1$predValues)^2)
mean((data2$y - pred2$predValues)^2)
mean((data2$y - pred3$predValues)^2)