Meta-analyses: Heterogeneity can be a good thing

• Published in: Past, Present, and Future of Statistical Science pp. 405 - 414
• Format: Book Chapter
• Language: English
• Published: Chapman and Hall/CRC 2014
• Subjects: Pre-existing Samples; Repeat Test Takers; ETS; Random Effects Approach; Specific Work Force; SAT; Untoward Side Effects; Framingham Heart Study; General Population Samples; Experimenter Expectancy Effects; Math Aptitude; Genetic Markers; Random Effects Analyses; Secondary Endpoints; Random Effects Method; General Population Cohort; Random Effects Model; Fixed Effects Approach; Genome Wide Association Scans; Term Meta-analysis; Find Disease Genes; Random Effects Meta-analysis; Laird Method; Genetic Epidemiology
• DOI: 10.1201/b16720-41

Meta-analysis seeks to summarize the results of a number of different studies on a common topic. It is widely used to address important and dispirit problems in public health and medicine. Heterogeneity in the results of different studies is common. Sometimes perceived heterogeneity is a motivation for the use of meta-analysis in order to understand and reconcile differences. In other cases the presence of heterogeneity is regarded as a reason not to summarize results. An important role for meta-analysis is the determination of design and analysis factors that influence the outcome of studies. Here I review some of the controversies surrounding the use of meta-analysis in public health and my own experience in the field.