158 – Modeling Response for Tailoring in Contact Strategies in Censuses and Surveys
Using Integer Programming to Determine the Optimal Assignment of Strategies for Mode Switching from Internet to Mail
John Chesnut
U.S. Census Bureau
With many surveys now embracing adaptive survey design methods, new challenges have arisen in applying optimization methods to meet the requirement to produce automated real-time decisions to adapt survey design strategies across sample cases simultaneously. Chesnut (2013) describes a mode-switching process for switching sample cases from Internet to mail using integer programming with the objective to maximize timeliness while controlling cost, response, and sample representativity. Building on this work, this paper presents the application of an integer programming solution using the opt model procedure in SAS� to automate the mode-switch decision process. We discuss the use of indicator variables to enable linear representations of our objective function for maximizing timeliness and the constraints for cost, response, and sample representativity in the integer programming environment. While this allows for a complete representation of the optimization problem at hand, the numerous constraints pose a challenge for computing a solution. We discuss alternative linear representations of our constraints using linear approximation methods to enable computed solutions.