The planning of Preliminary Sample: methodological aspects and an application to the Italian Monthly Retail Trade Survey

The standard approaches for dealing with provisional estimates essentially focus on the study and the definition of efficient estimator, exploiting the auxiliary information almost exclusively in the estimation phase. In this paper, we propose an extended appliance of the auxiliary information, developing an overall strategy involving both the sample design and the estimator. The work aims to define a preliminary sub-sample of units included in the planned sample for final estimates. Afterwards, the provisional estimates are computed by means of an appropriate estimator applied to the preliminary sub-sample data. In order to design the preliminary sub-sample, the balanced sampling theory is exploited. In the longitudinal survey context, where timeliness is a pressing target, the approach represents an approximately bias-robust sampling strategy. Results of a real survey data simulation study comparing the proposed strategies with strategies ignoring the planning of the preliminary sub-sample are presented.