Meta-analysis in ecology: Retrospective, prospective, and perspective

Meta-analysis was introduced to ecology in the early 1990s, borrowing on work from the social sciences and medicine but quickly developing statistical techniques and methodology more closely suited to the specific needs of this discipline. Publication of ecological meta-analyses has increased roughly exponentially since that time. We consider the nature of some of these ecology-specific developments in meta-analysis including introduction of new metrics of effect size, software, and development of resampling, factorial and mixed model approaches for ecological meta-analyses. We evaluate some of the issues that have hindered development and application of rigorous quantitative research synthesis in ecology, and those that have facilitated it.  Difficulties include differences between the nature of ecological and medical data, objections to quantitative data synthesis, lack of training in meta-analytic methods, continuing acceptance of discredited statistical approaches by authors, reviewers and editors, and publication and research bias. Areas in which new quantitative methods are being developed or need development include broadening the range of effect size measures, developing methods to account for non-independence (e.g. phylogenetic relatedness), and Bayesian and multivariate meta-analysis for ecological applications. We review several areas of successful application of meta-analysis in ecology including testing ecological theory, synthesizing responses to global climate change, and combining results from multi-site experiments, and highlight research areas (e.g. conservation biology) that can potentially particularly benefit from application of meta-analytic techniques.