Using Weights in the Analysis of the KFS Data: KFS Professional Development Workshops

Abstract

Although researchers using a complex sample survey data need not understand the in-depth details of the process generating the survey weights, questions about why, when, and how to use weighting and which statistical packages to use are the most common during data analysis. Unfortunately, relatively little practical information is available about why, when, and how to use weights, which makes weighting a challenging subject in the analysis of the KFS data. As noted by Kish (1992), one cannot find textbooks or references that provide a reasonable and comprehensive explanation of the use of weights in complex sample data. Although the theory of design-based inference from probability sample design has been under development for the last hundred years and is still an evolving field, the question to weight or not to weight in driving statistical models is an ongoing debate; however, the consensus among design-based analysts is that using weights in the statistical analysis of complex sample survey is never a wrong thing to do. This paper is aimed at clarifying the complex design of the KFS, thereby providing a simple guide to weighting and introducing the statistical procedures by the major statistical packages that are designed for analyzing data derived from a complex sample survey.