Meta-analysis is becoming increasingly popular in the field of ecology and environmental management. It increases the effective power of analyses relative to single studies, and allows researchers to investigate effect modifiers and sources of heterogeneity that could not be easily examined within single studies.
Many systematic reviewers will set out to conduct a meta-analysis as part of their synthesis, but meta-analysis requires a niche set of skills that are not widely held by the environmental research community. Each step in the process of carrying out a meta-analysis requires decisions that have both scientific and statistical implications.
Reviewers are likely to be faced with a plethora of decisions over which effect size to choose, how to calculate variances, and how to build statistical models. Some of these decisions may be simple based on appropriateness of the options. At other times, reviewers must choose between equally valid approaches given the information available to them. This presents a significant problem when reviewers are attempting to conduct a reliable synthesis, such as a systematic review, where subjectivity is minimized and all decisions are documented and justified transparently.
The authors propose three urgent, necessary developments within the evidence synthesis community. Firstly, they call on quantitative synthesis experts to improve guidance on how to prepare data for quantitative synthesis, providing explicit detail to support systematic reviewers. Secondly, they call on journal editors and evidence synthesis coordinating bodies (e.g. CEE) to ensure that quantitative synthesis methods are adequately reported in a transparent and repeatable manner in published systematic reviews. Finally, they propose that where faced with two or more broadly equally valid alternative methods or actions, reviewers should conduct multiple analyses, presenting all options, and discussing the implications of the different analytical approaches.
The authors argue that it is vital to tackle the possible subjectivity in quantitative synthesis described herein to ensure that the extensive efforts expended in producing systematic reviews and other evidence synthesis products is not wasted because of a lack of rigour or reliability in the final synthesis step.