Seminar | A comparison of two classic and two newer methods for meta-analysis: are they equally appropriate for systematic review?

Abstract

Meta-analysis methods are traditionally divided into the fixed effect approach and the random effects approach.  In 2014 Doi et al. argued for a hybrid “inverse variance heterogeneity” (IV-Het) approach.  In 2018 Rice et al. proposed a new approach with a subtly different name: fixed effects (note the plural). Here, I illustrate their different interpretations, and properties.  Known limitations and strengths of the classic fixed effect and random effects models are confirmed while IV-Het combines aspects of both.  The fixed effects (plural) method emerges as profoundly different: researchers planning systematic reviews should take note before specifying their analysis method.

Bio

Richard Stevens has been a medical statistician in diabetes research, cancer epidemiology and primary care research for over twenty years.  Since 2017 he has been course director for the MSc in EBHC Medical Statistics: a part-time medical statistics course that is designed for health professionals rather than mathematicians.

Zoom Link: Click here to join the meeting
If prompted for a password, please enter: 382071