Abstract
This article presents a synthesis of several widely used methods of estimation in survey sampling, including post-stratified estimation, regression estimation, and calibration estimation or generalized raking. All of these methods come under the general formulation of calibration estimation, and all of them are based on post-stratification given categorical auxiliary variables. Indeed, post-stratification is the finest calibration, and calibration is the relaxed post-stratification. Some results derived from such a perspective enable us to bring conditional inference of to stratified simple random sampling by means of Holt and Smith calibration estimation.