Estimation in multiple frames surveys

Abstract

In classic finite population sampling a basic hypothesis is the availability of a unique and complete list of units forming the target population to be used as a sampling frame. One often finds that a frame known to cover approximately all units in the population is one in which sampling is costly while other frames (e.g. special lists of units) are available for cheaper sampling methods. However, the latter usually only cover an unknown or only approximately known fraction of the population. In some cases a set of two or more lists is available for survey purposes. The general case of several lists, singularly partial and possibly overlapping, is known as Multiple Frame Survey. Multiple-frame surveys have been studied by several authors, with primary focus on point estimation (see Hartley 1962, 1974; Fuller and Burmeister 1972; Bankier 1986; Kalton and Anderson 1986; Skinner 1991; Skinner and Rao 1996, Rao and Wu 2010, Metcalf and Scott, 2009,… ). In this talk we present a calibration approach to inference from multiple-frame surveys. Our proposed approach addresses both point estimation and confidence intervals, and known auxiliary population information can be conveniently incorporated into inferences through constrained minimization. The theoretical properties of the proposed estimators are sketched and simulations studies are conducted to illustrate their finite size sample performance. Finally we present the package Frames2 for point and confidencial estimation in dual frame surveys.