Structural Equation Model (Structural Equation Model) is a model method for establishing, estimating and testing the causal relationship between multiple variables in a research system. It can replace multiple regression, factor analysis, covariance analysis and other methods, and it can be clearly displayed in a graphical model Studying the causal network relationship between variables in the system is a statistical method widely used in the fields of geoscience, ecology, evolution, environment, medicine, society, and economy in recent years. However, since Wright proposed the first path/path (Path Analysis) analysis (that is, the structural model in the structural equation model) method in the Proceedings of the National Academy of Sciences (PNAS) in 1920, the structural equation model has been developed for more than 100 years. A relatively large theoretical system and complex and changeable forms have been developed, which often make beginners at a loss. This course will use the open source software R platform to focus on ecological research issues as the main line, select a large number of classic cases, and systematically introduce the establishment and fitting of structural equation models through a combination of theoretical explanation and practical operation. The whole process of , evaluation, screening and result presentation enables students to use the structural equation modeling method to solve relevant scientific problems encountered in actual research and work. This course includes 8 topics, including introduction to R language and introduction to the principles of structural equation modeling (see the course content introduction), which is suitable for beginners of R language and structural equation modeling, and also suitable for graduate students and graduate students who have advanced application requirements for structural equation modeling. researcher.
Topic 1: Unified foundation: complete learning before class [Introduction to R and Rstudio and Structural Equation Modeling (SEM) Ecological Field Application]
Ai Shang training provides teaching materials + tutor assistance
1) Introduction to R and Rstudio: background, software and package installation, basic settings, etc.
2) Basic operation of R language, including vector, matrix, data frame and data list generation and data extraction, etc.
3) R language data file reading, sorting (cleaning), result storage, etc. (including tidverse)
4) R language basic drawing (including ggplot): basic drawing, typesetting, publication quality drawing output storage
1) Definition of SEM, application in the field of ecology and historical review
2) Basic structure of SEM
3) SEM estimation method
4) Path rules of SEM
5) Meaning of SEM path parameters
6) SEM analysis sample size and model identifiable rules
7) Basic process of SEM construction
Topic 2: Introduction to SEM analysis in R language: lavaan VS piecewiseSEM
1) Introduction to the application of structural equation modeling in ecological research and a review of key points of the model
2) Structural square model estimation method: basic working principle, main difference and application scenario analysis of local estimation and global estimation
3) The direct and indirect effects of species richness restoration in the case community (direct and indirect effects): basic process of SEM analysis-lavaan vs piecwiseSEM
(1) Model establishment
(2) Model fitting
(3) Model Evaluation
(4) Result display
Exercises after class: 1. Build a model based on the meta-model
Exercises after class: 2. Analysis of direct, indirect and moderating effects of plant community recovery after fire disturbance
Topic 3: Advanced application of lavaan-based SEM in the field of ecology
Case 1: Analysis of the direct and indirect effects of wetland ecosystem primary productivity
(1) Question formulation, meta-model construction
(2) Model construction and model estimation
(3) Model evaluation: path addition and deletion principles, optimal model screening method
(4) Result expression
Case 2: Evaluation of Plant Community Restoration Effects After Fire Disturbance - Missing Data and Insufficient Normality Data Processing -
Case 3: Analysis of the impact of grazing on the relationship between altitude and biomass - data group analysis
Case 4: The impact of the proportion of agricultural land on the abundance of water and grass in the estuary - data hierarchical/nested analysis
After-class exercise: Effects of environmental heterogeneity and resource availability on understory vascular plant diversity at different successional stages
Topic 4: Application of lavaan-based SEM latent variable analysis in the field of ecology
(1) Definition, advantages and application background analysis of latent variables
(2) The basic principle of lavaan implementation of latent variable analysis
(3) Case 1: Prediction of Ecological Restoration Performance of Spartina Community in Coastal Zone - Construction of Single Latent Variable Model
(4) Case 2: The direct and indirect effects of land use on flowering plant resources and visiting insects in urban landscapes - construction of multiple latent variable models
After-class exercise case: the impact of plant diversity, energy gradient and environmental gradient on the pattern of animal diversity-constructing latent variables of animal diversity
Topic 5: Application of lavaan-based SEM composite variable analysis in the field of ecology
(1) Definition of compound variable and application scenario analysis in the field of ecology
(2) Composite variable analysis lavaan implementation approach
(3) Case 1: Analysis of the formation mechanism of ecological force and biodiversity - the construction of multiple compound variables of soil physical and chemical factors
(4) Case 2: The impact of vegetation restoration on species richness after fire - composite variables to solve nonlinear problems
(5) Case 3: Compound impacts of climate warming and sea level rise on wetland plant communities - compound variables to solve interaction problems
After-class example explanation: Does plant community species diversity increase its resistance to invasive plants (Oikos, 2017) - multi-composite variable realization
Topic 6: Local Estimation SEM -piecewiseSEM and advanced applications in the field of ecology
(1) piecewiseSEM analysis of binomial and Poisson distribution data for endogenous variables
(2) Case 1: The impact of climate fluctuations on the food web structure of seagrass bed ecosystems - the impact of data layering and nesting, temporal and spatial autocorrelation on the results
(3) Case 2: The impact of species attributes and social evolution characteristics on the range and abundance of marine shrimp - phylogenetic corrections
(4) Case 3-5: PiecewiseSEM realization of grouped data, interaction, nonlinear relationship, etc. (the example data is the same as Topic 3)
After-class exercise case: Human activities, environmental conditions, and species attributes contribute to the relative contribution of animal domain size - grouping analysis and categorical variable processing
Topic 7: Application of Bayesian SEM in the field of ecology
(1) Introduction to Bayesian method
(2) Introduction of R language Bayesian SEM implementation package blavaan and brms
(3) Case 1: The impact of climate and niche overlap on the species richness of voles: model comparison, calculation of direct and indirect effects (blavaan)
(4) Case 2: Factors affecting vegetation restoration after fire - model fitting, model comparison and evaluation (brms)
Example for homework: The impact of biogeographic history on the primary productivity of forests in the northern hemisphere (BRMS)
