Using Linear Programming to Design Samples for a Complex Survey
Purchase Print Copy
|Add to Cart||Paperback56 pages||$20.00||$16.00 20% Web Discount|
RAND was asked to design and select a sample of Air Force personnel who would be asked to participate in an online survey of cultural attitudes. The design needed to minimize the number of people asked to participate so as to reduce the survey burden on a population already frequently invited to take surveys; reflect the response rates anticipated from previous surveys of the population; ensure adequate representation of a number of minorities of interest (rank, job type, race and ethnicity, gender, religion, and component); sample enough people in each of the overlapping subset categories of interest (e.g., black female NCOs) to allow for statistically meaningful comparisons; and minimize the number of service members invited to take both this survey and a health survey on an overlapping set of topics scheduled for the same time period. This report describes the method developed for designing joint samples for both surveys.
Table of Contents
Constraints on Numbers of People Assigned to Surveys and Their Sections
Constraints to Ensure Adequate Precision of Estimates
Tradeoffs Among Samples of Different Designs