New Resampling Method for Two-Phase Sampling Designs

• Activity Number: 36
• Type: Contributed
• Date/Time: Sunday, July 31, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Erika Antal*+ and Yves Tillé
• Companies: University of Neuchâtel and University of Neuchâtel
• Keywords: variance-estimation ; bootstrap ; two-phase

Abstract

In complex sampling designs a large part of existing bootstrap methods use simple random sampling with replacement in order to provide an unbiased estimator of the variance. When these methods do not take the sampling design into account, they provide biased variance-estimator. Resampled units usually need to be rescaled or weighted to correct this bias. Another set of methods consists of constructing artificial populations and to resample from them. These methods are very often time-consuming and have rounding problems. Moreover, when the sampling design is made of several phases, the implementation of these bootstrap methods becomes very difficult. We present new resampling method for two-phase designs that consists of resampling only a subsample of the units in the second phase. This subsample is selected randomly in such a way that it directly reproduces the appropriate variance, without rescaling or artificial population. The main advantage of the method is its simplicity as the after treatments like calibration or imputation for nonresponse can be directly applied on the bootstrap samples. That is why the proposed method is particularly worthwhile in real applications.