ShrinkageEstimates: Shrinkage Estimation of Small Domain Parameters

In nchenderson/shrinkcbp: CBP Shrinkage Estimates of Small Domain Parameters

Description

Computes shrinkage estimates for small domain parameters using different choices for the regression weights.

Usage

ShrinkageEstimates(Y, X, sd.resid, weights=c("cure", "cure.plugin", 
                  "cure.multitau", "mle","bpe"), vc.est=c("mle", "reml","ure"))

Arguments

Y	The K x 1 vector of responses.
X	The K x p design matrix.
sd.resid	The K x 1 vector sampling standard deviations.
weights	The method by which to choose regression weights.
vc.est	When the mle method is used for regression weights, this determines which method is used to estimate the variance component. This can be either maximum likelihood estimation, restricted maximum likelihood estimation, or minimization of an unbiased risk estimate.

Value

A list with the following components

Author(s)

Nicholas Henderson

Examples

K <- 20
X <- cbind(rep(1, K), rnorm(K))
beta.coef <- rnorm(2)
nn <- 1 + rpois(K, lambda=4)
sig.sq <- 1/nn
ran.ef <- rnorm(K)

Y <- X%*%beta.coef + ran.ef + sqrt(sig.sq)*rnorm(K)
sd.resid <- sqrt(sig.sq)


cbp.est <- ShrinkageEstimates(Y, X, sd.resid, weights="cure")
eblup.est <- ShrinkageEstimates(Y, X, sd.resid, weights="mle")
obp.est <- ShrinkageEstimates(Y, X, sd.resid, weights="bpe")

par(mfrow=c(1,2))
plot(cbp.est$shrinkage.estimate, eblup.est$shrinkage.estimate)
plot(cbp.est$shrinkage.estimate, obp.est$shrinkage.estimate)
abline(0, 1)

nchenderson/shrinkcbp documentation built on March 11, 2021, 2:52 p.m.