A Two-phase Sampling Strategy for Estimating Multiple Mean Values in the Presence of Nonresponse

The phenomenon of nonresponse in sample surveys usually results in biased estimates of population characteristics. One of the means to deal with nonresponse is the subsampling technique. It relies on re-contacting some subset of nonrespondents by using more expensive and more efficient tools (e.g. direct interview) than those used in the first attempt to collect data. This allows to increase response rate and to obtain unbiased estimates of population characteristics. In this paper, the problem of establishing the sample and subsample sizes minimizing the expected cost of the survey, while achieving desired precision of multiple mean value estimates, is considered. An algorithm is proposed that allows to establish the optimum initial sample and subsample sizes for two-phase sampling strategy.