Robust Estimation in Stratified Sampling under Super-population Model

In this book, two estimators of population total in stratified sampling under the single super-population model across the strata were developed and its properties and robustness were studied. It has been shown that both estimators remained robust and BLUE even under the model for stratified balanced and overbalanced samples. It has also been found that both estimators were more efficient than that of due to Scott et al. (1978 and Royall & Herson (1973b) under certain condition. The main tool in development of linear predictors in stratified sampling was the results of Pereira and Rodrigues (1983), which have been briefly mentioned in the book. Following the work of Bouza (1990, 1994), two sets of shrunken estimators were also developed under the model M and Mh respectively, in stratified sampling, where slopes of the models were common and uncommon across the strata. The proposed shrunken estimators have been found more efficient and robust as compared to that of Bouza (1990, 1994).