Evaluation of Descriptive Analyses of Survey Variances

• Activity Number: 278
• Type: Contributed
• Date/Time: Wednesday, August 14, 2002 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods*
• Author(s): Donsig Jang*+ and John Eltinge
• Affiliation(s): Mathematica Policy Research, Inc. and U.S. Bureau of Labor Statistics
• Keywords: Average design effect ; Average standard error ; Complex sample design ; Generalized variance function ; Parametric inference; Subpopulation

Abstract

Sample survey organizations often characterize the precision of point estimators through approximations based on, e.g., average design effects, average standard errors, generalized variance functions, or average confidence interval widths. Practical evaluation of the adequacy of these approximations will depend on whether one is interested in: 1.) summary descriptions of a fixed set of variance estimates or confidence interval widths; or 2.) formal inference (e.g., confidence intervals or test statistics) for related functions of finite-population or superpopulation parameters (e.g., design effects or specific coefficients of GVF models). Following a review of issues 1.) and 2.), this paper uses fixed-effect analysis of variance methods to develop some specific diagnostics for issue 1.). Some of the proposed diagnostics are applied to data from the U.S. Third National Health and Nutrition Examination Survey (NHANES III).