TWO-PHASE SAMPLING WITH CALIBRATION WEIGHT ADJUSTMENT IN SMALL-AREA AREA ESTIMATION FOR LONGITUDINAL SURVEYS

Accurate estimation of population parameters in small area estimation (SAE) is crucial, particularly in double sampling, where non-response can introduce significant bias. Although traditional estimators have been widely used to improve estimation accuracy, they often fail to fully adjust for non-response bias in small domains. This study proposes a new calibration-based estimator that extends the framework of an existing estimator by incorporating a Kullback-Leibler distance function to adjust the design weights. The estimators are derived under two conditions: (1) non-response affecting only the study variable and (2) non-response affecting both the study and auxiliary variables. This study evaluates the estimator’s performance using real-life data from the HFCS and the IHS in Nigeria (2019-2021), analyzing three distinct time periods: before, during, and after COVID-19. The MSE analysis results confirm that the proposed estimator consistently outperforms traditional SAE estimators, particularly in high non-response scenarios. The findings of this study suggest that integrating calibration techniques into double sampling frameworks significantly enhances estimation efficiency. The proposed calibration estimator is a robust and computationally feasible approach for improving domain mean estimation in longitudinal surveys, particularly when non-response rates are high. This study demonstrates superior efficiency and reliability in two-phase sampling under stratified random sampling by comparing the proposed estimator with existing methods.