Professor Stony Brook University Stony Brook, NY, United States
Background/Question/Methods Thirty years ago I published my first meta-analysis. What’s changed since then? What can we anticipate next? It took several years for ecologists to recognize the fundamental change in thinking that meta-analysis offers, but these conceptual and statistical tools have now been taken up with enthusiasm and have been transformative to the science of ecology. Meta-analysis in ecology has developed both like and unlike other disciplines. In that 1992 paper I chose to approach broad, complex questions using statistical tools borrowed from the social and medical sciences both to try to answer those questions but also to bring the attention of ecologists to the potential for these powerful tools. Wide-ranging ecological meta-analyses trade gains in the ability to generalize and examine multiple hypothesized causal factors, for issues with heterogeneity and sometimes confusing or ambiguous results. Other fields approach research synthesis and systematic reviews far more narrowly and precisely, sacrificing generality and wider applicability for reduced heterogeneity. Despite many challenges, the discipline of ecological meta-analysis has become firmly established, and faces many opportunities for future refinement and development.
Results/Conclusions Much has changed since the first ecological meta-analyses. In 1992, there were no electronic scientific databases, and journals were available as printed volumes in university and personal libraries, not online. Meta-analysis more broadly has expanded to become part of research synthesis, with the development of systematic reviewing. Statistical methodology and computational power have advanced and become more powerful. Search and inclusion criteria have become far less biased and more rigorous. Unfortunately documentation standards, open science, and transparency have been adopted in ecological research synthesis more slowly and unevenly than in other fields. Methodological rigor is too frequently ignored. Meta-regression, literature searches using machine learning, network meta-analysis and meta-analysis of networks are exciting advances. Many important ecological meta-analyses have contributed to our understanding at larger scales. Open science for accessing primary data has helped….a little. Fundamental problems including inherent imbalance and bias in sources, non-independence remain almost intractable challenges. “Big data,” compilation, and analyses of large databases have become fashionable but ignore many of the lessons learned in statistical systematic reviews and meta-analysis. There is a compelling need for greater formal training for students so that meta-analyses can be critically carried out, reviewed and “consumed” by readers.