Statistical innovation accounts for evolutionary bias in meta-analysis

NESCent postdoctoral fellow Marc Lajeunesse has proposed a solution to the problem of evolutionary bias in meta-analysis.

Meta-analysis is a statistical tool institutionalized in the medical sciences to review and summarize the findings of multiple studies. In fact, passing the rigors of this analysis is often used as a litmus test to decide when research on drug development can move from theory to practice. Ecologists and evolutionary biologists also rely heavily on meta-analysis to test theory, but unlike medical research, studies are not continuously repeated among different patients. Here studies are repeated across different species.

Summarizing studies from different species is a challenge for meta-analysis because it requires a degree of independence among studies. However, the many similarities among closely related species can violate this requirement. For example, dogs are similar to wolves as squashes are to pumpkins, but a conventional meta-analysis would treat all these species as independent; despite the apparent shared ancestry (or non-independence) among the two groups.

In an upcoming paper in The American Naturalist, Lajeunesse addresses this issue by providing a solution to the problem of non-independence of species for meta-analysis. This solution allows for information on the shared evolutionary ancestry of species to be included in all the conventional statistics of meta-analysis. Further, he extends this solution to allow for different hypotheses on the influence of shared ancestry. This is important given that many key ecological principles may assume different forms of evolutionary phenomena among different groups of organisms.

This statistical innovation fills a gap in the tools needed to assess and account for evolutionary bias when synthesizing ecological research. These tools can also help address bias in the systematic appraisal of current conservation practices. Here a meta-analysis biased by the shared ancestry of species can have a severe effect on policy decisions; perhaps leading to management inaction when species and habitats need protection.

Lajeunesse's results will appear in the September 2009 issue of The American Naturalist.

Lajeunesse, M. A. (2009). "Meta-Analysis and the Comparative Phylogenetic Method." The American Naturalist 174(3): 369-381. [ more ]