Seminar Series: Lisa Avery

Date: March 12
Time: 1:30 p.m. EST

"An overview of respondent-driven sampling: current projects and statistical methods"

Lisa-Avery.jpgLisa Avery

Senior Biostatistician
University Health Networks


Background: Respondent driven sampling (RDS) is a relatively new technique used to recruit participants from difficult to reach populations.  A number of studies are currently underway in Ontario which employ RDS including: the Our Health Counts studies of urban Indigenous health, the EMPOWER study of healthcare workers and, a COVID-19 study of children with complex needs. There are two aspects of RDS datasets that make undertaking regression analysis problematic: the non-equal sampling probability of the participants and, the dependency between observations.

Methods: Simulation studies were performed to 1) evaluate the validity of various regression models that could control for these two aspects of the data and 2) to evaluate the robustness of a new prevalence estimator.

Results: Type-I error rates were unacceptably high for weighted regression models, dependency within the data was in general inconsequential. Even when reported degree is accurate, as in this simulation, low reported degree can unduly influence regression estimates. The homophily configuration graph estimator is robust and less biased than other estimators.

Summary: Based on the simulation results, unweighted regression should be used with RDS data and sample clustering can be ignored, at least under conditions of moderate homophily. The new homophily configuration graph estimator is the preferred estimator for disease prevalence.

Zoom link:
Meeting ID: 959 1396 2951
Passcode: epibio

Dial-In by Phone
+1 647 558 0588 (Canada)
Meeting ID: 959 1396 2951
Passcode: 323551