Practical interpretation of ecological meta-analyses

Abstract

Meta-analysis offers ecologists a powerful tool for knowledge synthesis. There is however obvious and subtle issues to consider specific to ecology with respect to the appropriate interpretation of meta-analyses once the statistics are completed. Meta-analysts in any field must clearly define a priori the scope of inference and the purpose of the meta-analysis, but ecological meta-analysis often faces issues particular to this field. For example, the primary studies being combined, even for tests of the same hypothesis, in ecology are generally conducted at different study sites or have difference ecological contexts (i.e. often few studies are not conducted in a laboratory) and use very different methods to test a single hypothesis. General objectives of the meta-analysis could include assessment of confidence limits of the treatment or ecological process of interest, detection of differences in treatments from no effect, identification of research gaps, and differences between groups such as different populations or communities.Meta-analyses can also be used in ecology to assess whether the scope of tests of a hypothesis are adequate. Reporting more than one summary statistic such as different effect size metrics is an excellent means to enhance synthetic potential for ecologists. Magnitude and sign of effect sizes can also be interpreted in novel ways for ecology given the broad scope of forms of hypotheses explored in this discipline. Ecology is now poised to take advantage of the synthesis developments common in other disciplines and this brief conceptual methods review provides the appropriate framing for this endeavor.

Data-based, synthesis-driven: setting the agenda for computational ecology

“…These are often purely statistical, in that they aggregate signi cance and e ect size, to measure how robust a result is across di erent systems. Meta-analyses most often require a critical mass of pre-existing papers (Lortie et al 2013). Although they are irreplaceable as a tool to measure the strength of results, they are limited by their need for primary literature with experimental designs that are similar enough.Predator-preys (and other biotic) interactions have been studied with a few computational approaches to date.…”