Empirical likelihood confidence intervals for complex sampling designs

Berger, Y.G. and De La Riva Torres, O. (2012) Empirical likelihood confidence intervals for complex sampling designs. Southampton, GB, Southampton Statistical Sciences Research Institute, 31pp. (S3RI Methodology Working Papers). (Submitted)


PDF - Pre print
Download (227Kb) | Preview


There are little theoretical evidences that the central limit theorem holds for complex parameters under complex sampling designs. When it does not hold, standard confidence intervals may have poor coverages. The proposed approach gives design-based confidence intervals which may have better coverages than standard confidence intervals, pseudo empirical likelihood and bootstrap confidence intervals. The proposed approach does not rely on the normality of the point estimator, on variance estimates, design-effects, re-sampling, joint-inclusion probabilities and linearisation, even when the parameter of interest is not linear and biased. It can be used to construct confidence intervals for a large class of complex estimators and complex sampling designs. Our proposed approach can be viewed as an extension of the empirical likelihood approach to complex survey data. The proposed approach is simpler to implement than the pseudo empirical likelihood and the boostrap approaches. It also provides optimal point estimators. We apply the proposed approach to a measure of poverty based upon the European Union Survey on Income and Living Conditions (EU-SILC).

Item Type: Monograph (Working Paper)
Keywords: calibration, design-based approach, finite population corrections, h´ajek estimator, horvitz-thompson estimator, regression estimator, stratification, unequal inclusion probabilities
Subjects: H Social Sciences > HA Statistics
Divisions: University Structure - Pre August 2011 > Southampton Statistical Sciences Research Institute
Faculty of Social and Human Sciences > Social Sciences > Social Statistics & Demography
ePrint ID: 337688
Date Deposited: 01 May 2012 15:25
Last Modified: 27 Mar 2014 20:21
Publisher: Southampton Statistical Sciences Research Institute
URI: http://eprints.soton.ac.uk/id/eprint/337688

Actions (login required)

View Item View Item