Abstract: In this paper, we develop a pseudo empirical likelihood approach to incorporating auxiliary information into estimates from complex surveys. In simple random sampling without replacement, the method reduces to the empirical likelihood approach of Chen and Qin (1993). We show that the method is asymptotically equivalent to a generalized regression estimator in the case of estimating a mean or population distribution function with known population means for a vector of auxiliary variables. We go on to investigate, in a simple case, the incorporation of more complex auxiliary information, and demonstrate the resulting increase in efficiency using the proposed approach both theoretically and through a limited simulation.
Key words and phrases: Calibration, generalized regression estimator, jackknife, optimal regression estimator.