Propensity Stratification with Auxiliary Data for Address-Based Sampling Frames
Jamie L. Ridenhour
RTI International
Joseph P. McMichael
RTI International
Auxiliary information appended to Address-Based Sampling (ABS) frames can be incorporated at the sample design stage to increase the likelihood of sampling addresses with target groups of interest; however, this information needs to be accurate and nonmissing. For much of the auxiliary data currently available the drawbacks are that its accuracy is unknown and it is missing for many addresses on the sample frame making it difficult to use for stratification. In this paper we discuss how we solve this problem by using data collected from a nationally representative household survey of youths ages 11-16 to create a propensity model for stratification. This propensity model is then applied to the address frame for a subsequent survey targeting the same age group in rural areas and stratifies those addresses by how likely they were to have members of the eligible population. We discuss the impact this approach had on survey efficiency, both in terms of field cost and variance.