Empirical Bayes approaches have often been applied to the problem of estimating small-area parameters. As a compromise between synthetic and direct survey estimators, an estimator based on an empirical Bayes procedure is not subject to the large bias that is sometimes associated with a synthetic estimator, nor is it as variable as a direct survey estimator. Although the point estimates perform very well, naive empirical Bayes confidence intervals tend to be too short to attain the desired coverage probability, since they fail to incorporate the uncertainty which results from having to estimate the prior distribution. Several alternative methodologies for interval estimation which correct for the deficiencies associated with the naive approach have been suggested. Laird and Louis (1987) proposed three types of bootstrap for correcting naive empirical Bayes confidence intervals. Calling the methodology of Laird and Louis (1987) an unconditional bias-corrected naive approach, Carlin and Gelfand (1991) suggested a modification to the Type III parametric bootstrap which corrects for bias in the naive intervals by conditioning on the data. Here we empirically evaluate the Type II and Type III bootstrap proposed by Laird and Louis, as well as the modification suggested by Carlin and Gelfand (1991), with the objective of examining coverage properties of empirical Bayes confidence intervals for small-area proportions.
On a souvent utilise des approches empiriques Bayesiennes pour resoudre le probleme d'estimer des parametres d'aire petite. En tant que compromis entre les estimateurs d'enquete synthetique et direct, un estimateur fonde sur une procedure empirique Bayesienne n'est ni sujet au biais important souvent associe aux estimateurs synthetiques, ni aussi variable qu'un estimateur d'enquete direct. Bien que les estimateurs ponctuels aient une tres bonne performance, les intervalles de confiance naifs empiriques de Bayes tendent a etre trop courts pour atteindre la probabilite de couverture desiree, puisqu'ils n'incorporent pas l'incertitude creee par la necessite d'estimer la distribution a priori. Nous suggerons differentes methodologies d'estimation d'intervalle corrigeant les deficiences associees a l'approche naive. Laird et Louis (1987) ont propose trois types de bootstrap pour corriger les intervalles de confiance naifs empiriques de Bayes. Ayant denomme l'approche de Laird et Louis (1987) une approche inconditionnelle corrigeant le biais, Carlin et Gelfand (1991) ont suggere une modification au bootstrap parametrique de type III corrigeant le biais dans les intervalles naifs en conditionnant les donnees. Ici, nous evaluons empiriquement les bootstrap de type II et III proposes par Laird et Louis (1987), ainsi que la modification suggeree par Carlin et Gelfand (1991), dans le but d'examiner les proprietes de couverture des intervalles de confiance empiriques pour des proportions d'aire petite.
[1]
B. D. Riley,et al.
Iterative simulation methods
,
1990
.
[2]
N. Laird.
Empirical Bayes methods for two-way contingency tables
,
1978
.
[3]
R. Fay,et al.
Estimates of Income for Small Places: An Application of James-Stein Procedures to Census Data
,
1979
.
[4]
N. Breslow,et al.
Approximate inference in generalized linear mixed models
,
1993
.
[5]
R. Royall.
On finite population sampling theory under certain linear regression models
,
1970
.
[6]
G. Roberts,et al.
Logistic regression analysis of sample survey data
,
1987
.
[7]
David R. Cox,et al.
Prediction Intervals and Empirical Bayes Confidence Intervals
,
1975,
Journal of Applied Probability.
[8]
G. Y. Wong,et al.
The Hierarchical Logistic Regression Model for Multilevel Analysis
,
1985
.
[9]
T. Louis,et al.
Empirical Bayes methods for stabilizing incidence rates before mapping.
,
1994,
Epidemiology.
[10]
T. Louis,et al.
Empirical Bayes Confidence Intervals Based on Bootstrap Samples
,
1987
.
[11]
Bradley P. Carlin,et al.
A Sample Reuse Method for Accurate Parametric Empirical Bayes Confidence Intervals
,
1991
.
[12]
Noel A Cressie,et al.
REML estimation in empirical Bayes smoothing of census undercount
,
1992
.
[13]
Bradley P. Carlin,et al.
Approaches for Empirical Bayes Confidence Intervals
,
1990
.
[14]
N. Laird.
Nonparametric Maximum Likelihood Estimation of a Mixing Distribution
,
1978
.
[15]
Malay Ghosh,et al.
Bayesian Prediction in Linear Models: Applications to Small Area Estimation
,
1991
.
[16]
Scott L. Zeger,et al.
Generalized linear models with random e ects: a Gibbs sampling approach
,
1991
.
[17]
W. Keller,et al.
Disclosure control of microdata
,
1990
.
[18]
Thomas J. Tomberlin,et al.
PREDICTING ACCIDENT FREQUENCIES FOR DRIVERS CLASSIFIED BY TWO FACTORS
,
1988
.
[19]
T. Louis,et al.
A constrained empirical Bayes estimator for incidence rates in areas with small populations.
,
1994,
Statistics in medicine.
[20]
C. Morris.
Parametric Empirical Bayes Inference: Theory and Applications
,
1983
.
[21]
T A Louis,et al.
Random effects models with non-parametric priors.
,
1992,
Statistics in medicine.
[22]
B. MacGibbon,et al.
Protection against outliers in empirical bayes estimation
,
1994
.
[23]
Paul Neuhaus,et al.
United States Bureau of the Census
,
1998
.
[24]
Malay Ghosh,et al.
Small Area Estimation: An Appraisal
,
1994
.
[25]
L. Tierney,et al.
Accurate Approximations for Posterior Moments and Marginal Densities
,
1986
.