EBLUP: EBLUP estimator for small areas

In becarioprecario/SAE2: Small Area Estimation with R

Description

This function compute the EBLUP estimator for small areas (Rao, 2003) using a area level model (see references below). Either Maximum Likelihood (ML) or Restricted (REML) can be used in the estimation.

You can find more information and examples in the vignette included in the package.

Usage

EBLUP(formula, varformula, data, Z=NULL, tol=10e-5, maxiter=50, method="ML", na.action=NULL)
SEBLUP(formula, varformula, data, Z=NULL, W, tol=10e-5, maxiter=50, method="ML", na.action=NULL)

Arguments

formula	Formula relating the response to the covariates in the fixed effects.
varformula	A one sided formula indicating the variable which contains thesampling variance.
data	A data frame containing the variables used in the model.
Z	Structure of the random effects. By default, the identity matrix is used.
W	Adjacency matrix used in the Spatial EBLUP.
tol	Tolerance used in the computations.
maxiter	Maximum number of iterations of the fitting algorithm.
method	Either "ML" or "REML"

.

na.action

Action to handle missing values. At the moment, nothing is implemented.

Value

EBLUP	EBLUP estimates of the area values.
beta	Estimates of the coefficients of the fixed part of the model.
sigma2u	Estimate of the variance of the random effects.
g1	Estimate of first component (G1) of the MSE.
g2	Estimate of second component (G2) of the MSE.
g3	Estimate of third component (G3) of the MSE.
mse	Estimate of the MSE.
randeff	Estimates of the random effects.
varbeta	Estimate of the variance of the coefficients of the fixed part.

References

JNK Rao (2003).Small Area Estimation. John Wiley & Sons, Inc., Hoboken, New Jersey.

See Also

SEBLUP

Examples

#Load data
data(seblup)

spam.options(eps=.0000000001)

d<-data.frame(direst=ydir1, covariate=Xpop[,2], desvar=vardir)

eblupml<-EBLUP(direst~covariate, ~desvar, d)
eblupreml<-EBLUP(direst~covariate, ~desvar, d, method="REML")

summary(eblupreml$randeff - eblupml$randeff)

plot(eblupreml$randeff , eblupml$randeff)
abline(0,1)


plot(eblupml$mse, eblupreml$mse)
abline(0,1)

#Spatial EBLUP
seblupml<-SEBLUP(direst~covariate, ~desvar, d, W=W)
seblupreml<-SEBLUP(direst~covariate, ~desvar, d, W=W, method="REML")

becarioprecario/SAE2 documentation built on May 11, 2017, 3:20 a.m.