Composite-SEM for jamovi
Composite-SEM is a comprehensive module for jamovi designed to estimate, analyze, test, and study linear and multi-group structural equation models. Built upon the robust cSEM R package, this module brings professional-grade composite-based Structural Equation Modeling (SEM) to a user-friendly graphical interface.
Core Features
Advanced Estimation Techniques: Supports Partial Least Squares (PLS), Generalized Structured Component Analysis (GSCA), and Maxvar.
Flexible Measurement Models: Easily define Reflective (Latent) and Formative (Composite) constructs.
Structural Model Flexibility: Support for both Path models (user-specified relations) and Auto-Correlation models (Factorial weighting).
Comprehensive Diagnostics: Produces model fit indices, confidence intervals via bootstrapping, and out-of-sample prediction measures.
Multigroup Analysis (MGD): Perform difference tests across groups using permutation-based methods.
Linear Benchmarking: Compare PLS model performance against standard linear model (LM) benchmarks for predictive assessment.
Interface Overview
- Measurement Model Setup The interface allows for intuitive "drag-and-drop" variable assignment to define your model's building blocks.
Latent Variables: Define reflective constructs where indicators are caused by the latent factor.
Composite Variables: Define formative constructs where indicators form the composite proxy.
Multigroup Analysis: Assign a grouping variable to compare path coefficients and weights across different segments of your data.
- Model Estimation & Structural Relations Fine-tune your analysis through advanced estimation settings and text-based structural syntax.
Estimation Method: Choose between PLS, GSCA, or MAXVAR.
Bootstrapping: Enable resampling to calculate stable confidence intervals and p-values.
Structural Relations: Define paths using standard syntax:
Y ~ X for direct effects.
Y ~~ X for correlations/covariances.
Predictive Assessment: Toggle the Linear Model Benchmark to assess the out-of-sample predictive power of your composite model.