413 – Surveys of Children, Students, and Schools
Optimal Sampling Designs for National Surveys of Students
Kate Flint
ICF International
Ronaldo Iachan
ICF Macro
WIlliam Robb
ICF International
This article examines alternative sampling designs for national surveys of students. These surveys often seek to oversample demographic subgroups of students, such as African Americans or Hispanics, in order to generate precise estimates for these subgroups. This paper explores multistage stratified sampling designs with primary sampling units (PSUs) defined geographically to provide cost-effective data collection. These designs typically select PSUs with probabilities proportional to size (PPS) using a measure of size (MOS) based on functions of student enrollment, with the MOS calibrated so that the sample yields the required number of students by demographic subgroup. The canonical design uses student enrollment. Using a sampling frame constructed for national surveys, we will generate simulated samples under a variety of sampling designs. Designs will be evaluated using several metrics.
