Abstract:
|
Managing a rolling household panel leads to a natural turnover due to expiration or attrition. As such, an effective recruiting process is necessary to maintain sample size. For a small number of controls (demographics, geography, etc.) the recruiting can easily be achieved by a stratified sample design. When countless controls are necessary, an alternative solution would be to use a balanced sample design - a complicated procedure. The goal is to reduce the number of strata in order to make it easier to manage and retain a degree of flexibility.
In general, if there are N control variables (Age Group, Race/Ethnicity) each with M levels, then there are N x M possible strata available at the atomic level. In this instance, each household in the frame will belong to a single atomic stratum. The problem is to collapse some strata together to reduce the number of entities to X < N x M. This will be done in such a fashion so that a sample drawn from the new stratification which will closely approximate the N x M targets. The proper allocation for the new stratification will be addressed. We will demonstrate how this technique performs using Nielsen Radio Panels.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.