CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Cluster-based outcome-dependent sampling in resource-limited settings: Inference in small-samples Authors:  Sebastien Haneuse - Harvard TH Chan School of Public Health (United States) [presenting]
Sara Sauer - Harvard TH Chan School of Public Health (United States)
Bethany Hedt-Gauthier - Harvard Medical School (United States)
Claudia Rivera-Rodriguez - University of Auckland (New Zealand)
Abstract: Outcome-dependent sampling is an indispensable tool for carrying out cost-efficient research in resource-limited settings. One such sampling scheme is a cluster-based design where clusters of individuals (e.g. clinics) are selected, in part at least, on the basis of the outcome rate of the individuals. For a given dataset collected via a cluster-based outcome-dependent sampling scheme, it has previously been proposed to perform estimation for a marginal model using inverse-probability-weighted generalized estimating equations, where the cluster-specific weights are the inverse probability of the clinics' inclusion in the sample. We provide a detailed treatment of the asymptotic properties of this estimator, together with an explicit expression for the asymptotic variance and a corresponding estimator. Furthermore, motivated by a study we conducted in Rwanda, we provide expressions for small- sample bias corrections to both the point estimates and the standard error estimates. Through simulation, we show that applying these corrections when the number of clusters is small generally reduces the bias in the point estimates, and results in closer to nominal coverage. The proposed methods are illustrated using data from 18 health centers in Rwanda, collected via a cluster-based outcome-dependent sampling scheme, with the goal of examining risk factors for low birth weight.