Abstract: Structural equation modeling (SEM) is a major approach for analyzing ecological data. Differing from other multivariate statistical methods, SEM is mainly driven by assumptions and could quantify both direct and indirect causal relationships among multi-variables. Since the introduction of SEM into the field of ecology in China is relatively short, scientists encounter many problems when using SEM, with various mistakes. Therefore, we elaborated principles, processes, evaluation and modification of SEM in this review. With a case study, we introduced the usage of two major R packages for SEM, i.e. lavaan and piecewiseSEM. Concretely, lavaancan perform SEM with latent variables, while piecewiseSEM can incorporate nonindependent observations, and handle response variables with residuals not satisfying multivariate normal distribution. This review can help researchers understand and use SEM correctly and accurately, and promote the usage of SEM in ecology.

Key words: ecological complexity, causality, latent variable, path analysis, lavaan, piecewiseSEM.