Abstract: Recent development of the empirical likelihood method in survey sampling has attracted attention from survey statisticians. Practical considerations for using the method in real surveys depend largely on the availability of simple and efficient algorithms for computing the related weights for the maximum empirical likelihood estimators. In this article we briefly describe the modified Newton-Raphson procedure of Chen, Sitter and Wu (2002) for non-stratified sampling designs and show that under suitable reformulation the algorithm can also be used to handle stratified sampling, the most commonly used design in survey practice. The idea of the q-weighted empirical likelihood approach is briefly introduced and the related algorithms are discussed. The proposed algorithms are tested in a limited simulation study and are shown to perform well.
Key words and phrases: Newton-Raphson algorithm, pseudo empirical likelihood, stratified sampling, weighted empirical likelihood.