Prevalence meta-analysis Video 2: Statistical principles

In this second segment of the DeepDive, Dr. Ibraheem Abioye lays the statistical foundation for a rigorous prevalence meta-analysis. He explains the core concepts that govern how proportions are pooled across studies and why special considerations are required when analyzing prevalence data. Key topics include: - Purpose of meta-analysis and how pooled estimates increase precision across heterogeneous studies. - The logic of weighted averages, why larger studies contribute more, and how fixed-effects and random-effects models differ in their assumptions and interpretation. - Why random-effects models are the default for prevalence data and how between-study variance (τ²) is estimated using methods such as REML. - How case definitions, measurement tools, and cutoffs influence the true underlying prevalence. - The need for data transformations (raw, logit, double arcsine) when pooling proportions—especially for rare or extreme values—and guidance on choosing the right transformation for your dataset. - Preparing data in a wide format for analysis. By the end of this video, participants gain a clear conceptual understanding of the statistical mechanics of prevalence meta-analysis and are ready to implement these principles in R.