Allocation for Dual Frame Telephone Surveys with Nonresponse.

This paper presents several results for the optimal allocation of dual frame designs under different conditions, assuming the estimators are poststratified to domain totals. Both screener and overlap designs are discussed, and the results of the allocation for the two types of designs are illustrated by showing the efficiency of optimal designs in dual frame landline and cell phone surveys. We begin with results on optimal sample allocation to the frames assuming complete response and a linear cost function. We then compare the variance for estimates from the optimal screening and overlap designs, and extend these results to the situation where response rates differ for the landline and cell phone domains. We find that a screening design achieves a smaller variance than an overlap design with the same total cost under many cost structures. [ABSTRACT FROM AUTHOR]