Supplementing Sub-Samples with Population Data via Calibration

• Activity Number: 126
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Elizabeth Malecha*+ and Sebastien Haneuse
• Companies: Harvard and Harvard School of Public Health
• Keywords: Auxiliary data ; Calibration ; Influence function ; Case-control study ; Aggregate data ; Ecological data

Abstract

To address scientific questions, researchers typically obtain and analyze data on a sub-sample of individual units from the population of interest. In many instances, researchers additionally have access to partial information on all individuals in the population. When individuals fall naturally into groupings, such as in geographic-based studies, the partial information may consist of group-level, aggregate data. The use of group-level data is appealing: not only is inexpensive group-level data readily available due to the increased prevalence of large databases, group-level data often lends itself to more efficient estimators through greater exposure variability due to differences in between-group exposure rates, thereby increasing statistical power. In the absence of individual-level data, group-level data can not be used to draw individual-level conclusions on its own without making untestable assumptions, which when unsatisfied subject estimators to several potential sources of `ecological bias'.

Ideally, researchers would supplement a sample of individual-level data with existing, accessible group-level data. We consider the use of methods developed in the survey sampli