WL15: Common Misunderstandings of Weights in Survey Studies and Beyond

Music City Center 

Weights play a critical role in statistical analyses, especially in survey and observational studies, where sampling and propensity score weights are employed to address cluster sampling and selection bias. However, many researchers, including our collaborators, continue to struggle with understanding different types of weights and misuse them in practice. This roundtable will address two key issues. First, I will explore some common misunderstandings about different types of weights by providing an overview and clarifying their conceptual and practical distinctions to improve their proper application in statistical analyses. Second, I will discuss problems encountered when applying the Mann-Whitney-Wilcoxon Rank Sum Test (MWWRST) to survey data using the R package "survey" and introduce a new method developed to address these issues.

Sampling weights

Model-based variance estimator

Semiparametric models

Inverse probability weighting

Mann-Whitney U test 

Survey Research Methods Section