Title
Efficient imputation-based predictive ratio-type estimation with sub-sampling of non-respondents
Speaker
M.C. Agrawal, Department of Statistics, Delhi University, India and Department of Statistics, University of Akron
Abstract
In the event of non-response, we exploit the usual predictive framework under fixed population set-up to construct four imputation-based predictive estimators of the population mean wherein auxiliary information, assumed to be known in full,is tapped. These four estimators are compared with ratio-type estimators in Rao(1986). It is shown that one of the proposed estimators, depending on conditions that involve the correlations between the survey variable and the auxiliary variable within the population in the non-response stratum, is preferable to the earlier ones.
The variances of these four estimators and the optimum sampling and sub-sampling fractions for a specified cost or mean squared error, are also derived. An empirical investigation of the relative performance of potentially competing estimators is also discussed.
