Indirect sampling in context of multiple frames

Abstract

Multiple frame design is a strategy that deals with the problem of under coverage of sampling frames, which consists in combining several frames in order to provide complete or nearly complete coverage of the target population. In most cases, the frames overlap causing a problem to estimate in what regards sample weights computation. Therefore, Indirect sampling can be an alternative approach to the classical sampling theory in dealing with the overlapping problem of sampling frames on survey estimates. In this thesis, the classical estimators of multiple frames sampling - Domain Membership estimator and Unit Multiplicity estimator – are translated to the context of indirect sampling. Additionally the Optimal Deville and Lavallée estimator is decoded to the context of multiple frames surveys. The purpose is to deduce a new class of indirect sampling estimators capable of being applied in multiple frames surveys, more specifically in the particular case of dual frame surveys. The new estimators are then compared with the indirect sampling of Optimal Deville and Lavallée estimator, under eight different scenarios of links between sampling frame and target population, in order to identify which of them is more efficient. Henceforth, both theoretical comparisons and comparisons by simulation reveal that the Unit Multiplicity estimator and the Optimal Deville and Lavallée estimator are equally competent and are more efficient than the Domain Membership estimator.