PLS-SEM

Description

Estimates relationships between latent constructs measured by survey items using a variance-based approach, suited to prediction-oriented models and modest sample sizes.

Structural equation modelling links latent constructs, such as trust, perceived usefulness or effort, measured by multiple survey items, in a path model. PLS-SEM estimates these models by iteratively maximising explained variance in the endogenous constructs, rather than reproducing the covariance matrix as CB-SEM does. This makes it robust with smaller samples, non-normal data and complex models, and suited to prediction-oriented research. Rigour comes from evaluating the measurement model (indicator loadings, composite reliability, average variance extracted, discriminant validity using HTMT) before the structural model (path coefficients with bootstrapped confidence intervals, R squared, effect sizes and out-of-sample predictive power with PLSpredict). Constructs should come from theory or qualitative work, and causal interpretation depends on design and stated assumptions.

Attributes

Loss type
Siloed
Effort
Weeks

Steps

  1. Specify the path model from theory or qualitative findings, ideally checked with a DAG, defining constructs and their indicators.
  2. Collect survey data with at least 3 items per reflective construct, using a sample size justified by power analysis (often 100 to 300 respondents).
  3. Assess the measurement model: loadings above 0.70, composite reliability between 0.70 and 0.95, AVE above 0.50 and HTMT below 0.85 to 0.90.
  4. Estimate the structural model and bootstrap (for example 5,000 resamples) to obtain confidence intervals for path coefficients.
  5. Evaluate explanatory and predictive power using R squared, f squared and PLSpredict, and test mediation or moderation if hypothesised.
  6. Report the model, diagnostics, effects and limitations, linking constructs back to qualitative evidence.

Sources

  • Principles & Practice of SEM (Kline)
  • Primer on PLS-SEM (Hair et al.)
  • DAGs+SEM Playbook (Gopinath)
  • factor_analyzer
  • semopy
  • pyreadstat (SPSS import)
  • scikit-learn (MICE)
  • pingouin (reliability)
  • pandas
  • The Inquiry Series, Vol III, Ch 49
  • The Inquiry Series, Ch 49 — Structural Equation Modeling — From Quotes to Coefficients
  • The Inquiry Series, C3 — Survey Pipeline in Python — From Raw Responses to Structural Equation Model
  • A Primer on PLS-SEM
  • SmartPLS
  • Principles and Practice of Structural Equation Modeling (Rex Kline)
  • A Primer on Partial Least Squares Structural Equation Modeling (Joseph Hair et al.)
  • DAGs+SEM Playbook (Vivek Gopinath)
  • Structural Equations with Latent Variables (Kenneth Bollen)
  • A Primer on PLS-SEM — Hair et al.

Also known as

  • Partial least squares structural equation modelling
  • PLS path modelling
  • Variance-based SEM
  • Partial Least Squares SEM
  • Structural equation model
  • Structural model
  • Path analysis with latent variables
  • SEM path model
  • PLS-SEM for Small-Sample UX Modelling

Categories

  • Analytics & Behaviour
  • Statistics & Modelling