Exploring Geographic Clustering Methods in National School-Based Student Survey Samples

• Activity Number: 413
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): WIlliam Robb*+ and Ronaldo Iachan and Randal ZuWallack and Kate Flint and Paul Martino
• Companies: ICF International and ICF Macro and ICF International and ICF International and ICF International
• Keywords: oversample ; GIS ; geographic clustering ; SES ; school-based Samples

Abstract

When conducting school-based student surveys at the national level, multi-stage sample designs are typically employed. These designs form county-based clusters as the first stage of sampling to support cost-effective data collection; it is cost-prohibitive to send staff to a sample of schools distributed randomly across the nation. To address this, schools are grouped into Primary Sampling Units consisting of geographically proximate groups of counties. We note that other design features tend to yield samples that are geographically clustered. These strategies, adopted to oversample sub-groups such as African American and Hispanic students, include stratification with disproportional allocation at the first stage, and the use of a weighted measure of size. This paper explores the possibility of developing sampling techniques that address both goals simultaneously. Can we use methods that oversample particular subgroups of students, and yield geographically proximate clusters of schools in support of efficient data collection without the explicit clustering of schools into design PSUs? The answer to this question is explored via GIS based analysis of simulated samples.