Model Assisted Regression Estimators for Estimating Population Total

摘要

“Model Assisted estimators” is a well known term that is used for the estimators that can be interpreted both in model and design based paradigms. Model Assisted estimators give assurance that in case of failure of regression model the estimator will remains asymptotically design unbiased. Calibration is a well known model assisted technique of estimating population parameters. It generates calibration weights that satisfy a calibration to benchmark constraints and have minimum distance from the sampling design weights. In this paper two new model assisted estimators are proposed that adjust calibration and design weights to get modified weights for each observation of the study variable. The comparison of the proposed estimators is made with classical calibration estimator both on theoretical and on numerical basis.