American Journal of Mathematics and Statistics 2014, 4(2): 65-71
DOI: 10.5923/j.ajms.20140402.03
Estimation for Domains in Stratified Sampling Design in
the Presence of Nonresponse
E. P. Clement
1,*
, G. A. Udofia
2
, E. I. Enang
2
1
Department of Mathematics and Statistics, University of Uyo, Uyo, Nigeria
2
Department of Mathematics, Statistics and Computer Science, University of Calabar, Calabar, Nigeria
Abstract An analytical approach for finding the best sampling design subject to a cost constraint is developed. We
consider stratified random sampling design when elements of the inclusion probabilities are not equal but are in same stratum
and proposed estimators of totals for domains of study under nonresponse in the context of calibration estimation. We derived
optimum stratum sample sizes for a given set of unit costs for the sample design and compared empirically the relative
performances of the proposed calibration estimators with a corresponding global estimator. Analysis and evaluation are
presented.
Keywords Calibration estimation, Domain estimation, GREG-estimator, Optimum allocation, Sampling design
1. Introduction
In sample survey, separate estimates of a parameter may
be required for subpopulations into which a population is
divided without separately sampling from these
subpopulations. Such subpopulations are called domains of
study [1]. The method of estimating the domain parameters
is called domain estimation.
[2] first considered in detail some of the problems
associated with the estimation of domain totals, means and
proportions in the case of a single-stage simple random
sampling. He noted that the variance of an estimator of a
domain parameter is increased by the fact that the number of
the domain elements, and hence the number of those
elements that can fall in a random sample of a fixed size, is
unknown before the start of the survey. [3] gave a derivation
of Yates’ results in multi-stage sampling. [3] paper is one of
the first attempts to unify the theory of domain estimation.
Hartley provided the theory for a number of sample designs
where domain estimation was of interest. His paper mostly
discussed estimations that did not make use of auxiliary
information. He did, however, consider the case of ratio
estimation where population totals were known for the
domains.
[4] extended Yates’ results to double sampling for
probability proportional to size (PPS) when information on
the size, X, of each sampling unit is unknown. [5] proposed
an empirical Bayes estimation of domain means under nested
* Corresponding author:
epclement@yahoo.com (E. P. Clement)
Published online at http://journal.sapub.org/ajms
Copyright © 2014 Scientific & Academic Publishing. All Rights Reserved
error linear regression model with measurement errors in the
covariates. The problem of allocation of resources when
domains of study are of primary interest is discussed by [6].
However, despite these vast extensions of Yates results,
the phenomenon of nonresponse and its problems in domain
estimation have not been adequately addressed. In many
human surveys, information is in most cases not obtained
from all the units in the survey even after some call-backs.
An estimate obtained from such incomplete data may be
misleading especially when the respondents differ from the
non-respondents because the estimate can be biased.
Nonresponse always exists when surveying human
populations as people hesitate to respond in surveys; and
increases notably while studying sensitive issues like family
size. Nonresponse as an aspect in almost every type of
sample survey creates problems for estimation which cannot
simply be eliminated by increasing sample size.
The phenomenon of nonresponse in a sample survey
reduces the precision of parameters estimates and increases
bias in estimates resulting in larger mean square error, thus
ultimately reducing their efficiency.
An important technique to address these problems is by
calibration. Calibration as a tool for reweighting for
nonresponse was first introduced by [7] for the estimation of
finite population characteristics like means, ratios and totals.
This calibration approach requires the formulation of
suitable auxiliary variables. The calibration approach
provides a unified treatment of the use of auxiliary
information in surveys with nonresponse. In the presence of
powerful auxiliary information, the calibration approach
meets the objectives of reducing both the sampling error and
the nonresponse error.
In survey sampling many authors, such as [7-11] defined