Speaker: Natallia Katenka from The University of Rhode Island
Abstract: The spillover effect (or indirect effect) exists if an individual’s treatment can affect the outcomes of others. People who inject drugs (PWID) are part of social, sexual, or drug use (i.e., HIV risk) networks and treatment of one individual can affect other members of these networks socially or biologically. The spillover effects in these networks can be evaluated and leveraged to improve public health interventions. Although recent developments allow the estimation of spillover effects in networks, a focus on study design methodology is needed. Adequately powered studies are critical to the development and success of scientific studies; otherwise, the study could be over- or under-powered and may result in scientific, ethical, and economic issues. New methods will provide researchers with tools to generate adequately powered sociometric network designs for evaluating the spillover effect. The purpose of this work is to evaluate the impact of different design parameters on the power of detecting spillover effect, including network size and features, effect size, variance, and intra-cluster correlation, and to develop methods for power and sample size calculations for spillover in network-based studies with non-randomized interventions. Preliminary work showed that the power increased with more nodes (i.e., individuals) or effect sizes with higher absolute values. Interestingly, the number of components (i.e., a connected subgraph that separated from any other node in the entire graph) did not have a monotonic effect on the power when the number of nodes was fixed.
Natallia will be on-site for the seminar in W2008, but the seminar will also be streamed through Zoom.
ZOOM
Meeting ID: 928 8613 3715
Passcode: 838290
