Understanding biodiversity effects on trophic interactions with a robust approach to path analysis

Summary: With its facility to assess causal mechanisms among multiple variables, the application of path analysis in medical, natural, and social sciences has become widespread. Of the many types of path analysis, structural equation modeling (SEM), including Bayesian applications of this method, has gained popularity. However, SEM remains constrained by biased estimates in the case of model misspecification, while Bayesian methods are limited by time consumption and computational requirements. Here, we propose a novel estimator utilizing robust estimating equations combined within a Bayesian framework to improve multilevel path analysis. We apply this method to an ecological trophic interaction case study that assessed the path effects of global plant diversity on the interactions of plants, invertebrate herbivores, and their natural enemies. Using a simulation study, we show that this new estimator is unbiased and more robust. Moreover, the computational time cost for the estimating procedure is reduced compared with multivariate Bayesian analysis. Science for society: Path analysis is a useful method that helps researchers understand how multiple factors are connected and how they influence each other. However, the increasing volume of data and the complexity of interactions among factors necessitate the development of high-accuracy and high-efficiency statistical methods to enhance our understanding of biological processes and social complexity. Here, we present a benchmark study using simulations and apply the proposed method to investigate biodiversity effects on trophic interactions in ecosystems.