Reply to: Indiscriminate data aggregation in ecological meta-analysis underestimates impacts of invasive species

Andrea Anton, Nathan R. Geraldi, Catherine E. Lovelock, Eugenia T. Apostolaki, Scott Bennett, Just Cebrian, Dorte Krause-Jensen, Nuria Marbà, Paulina Martinetto, John M. Pandolfi, Julia Santana-Garcon & Carlos M. Duarte
Nature Ecology & Evolution, (2020)

Reply to: Indiscriminate data aggregation in ecological meta-analysis underestimates impacts of invasive species

Keywords

Ecology, Evolution, Biodiversity, Invasive species

Abstract

REPLYING TO M. S. Thomsen Nature Ecology & Evolution https://doi.org/10.1038/s41559-020-1117-6 (2020)

Exotic species are a growing ecological threat, although their factual ecological effects are currently debated. Our global meta-analysis recently reported that marine exotic species have an overall significant, but modest, effect on native communities. Thomsen argues that the data in this meta-analysis were aggregated in a manner that might underestimate the ecological impacts of marine exotic species. Here we discuss the data aggregation methodologies proposed by Thomsen and evaluate, when possible, whether the findings meet the expectations.

Meta-analysis is a powerful statistical tool for synthesizing evidence across independent studies and is commonly used in ecology. It involves the aggregation of extensive datasets to perform analyses that aim to answer specific questions. Thomsen argues that the weak significant impacts of marine exotic species on native communities reported in our recent meta-analysis were potentially due to an indiscriminate aggregation of data. Thomsen stated that he demonstrates “that many of the reported weak impacts occurred because effects were aggregated and averaged across opposing ecological processes and methodological approaches”. However, his analyses selected single examples for each of the three instances where, in his view, the way we aggregated (analysed) data underestimated the impact of marine exotic species. We evaluate these three instances, but rather than focusing on single examples, we reran the analyses with the entire database (where possible) to statistically confirm or refute Thomsen’s hypotheses2.


Code

DOI: 10.1038/s41559-020-1118-5

Sources

Website PDF

See all publications 2020