Abstract:
|
When obtaining information on a population of interest, state and local entities are often interested in producing estimates for small geographic areas. These areas may be a single county or a collection of counties. One of the more inexpensive modes to collect these data are through a telephone survey. Given population shifts in the types of telephones used, in order to obtain a representative sample, a dual-frame incorporating landline and cellphone phone numbers is used. To best represent the population, at least half the sample should come from the cellphone frame. However, for cellphone numbers there is no perfect identifier for county or county group. In this paper, we present and compare different methods for targeting small areas in a dual-frame telephone survey. We discuss how the classification errors can be derived and incorporated into the allocation of the desired sample size. Data from the 2015 Ohio Medicaid Assessment Survey (OMAS) is used as illustration.
|