Partial Least Squares Structural Equation Modeling
Also known as: PLS-SEM, PLS path modeling
PLS-SEM is a variance-based approach to structural equation modeling developed by Herman Wold (1985) that estimates latent variable models by maximizing the variance explained in dependent variables. Unlike covariance-based SEM, PLS-SEM is particularly useful for exploratory research, small to medium samples, complex models with many constructs, and non-normal data.
Key highlights
- Small sample efficiency: works with samples as small as 30-50, while covariance-based SEM needs hundreds
- Prediction focus: optimizes variance explained, making it ideal for predictive models
- Handles complexity: manages models with many latent variables, mediation chains, and non-recursive paths
- Non-normal data: does not require multivariate normality, robust to violations
- Practical software: widely implemented in accessible tools (SmartPLS, ADANCO, R packages)
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Apply PLS-SEM when exploring relationships in complex models with many constructs, sample sizes are small to moderate (30-200), when prediction is prioritized over perfect fit, or when data violates normality assumptions. Ideal for business, social science, and exploratory research. Avoid PLS-SEM if the goal is rigorous model fit testing (use covariance-based SEM instead) or if your focus is purely confirmatory.
Strengths & limitations
- Small sample efficiency: works with samples as small as 30-50, while covariance-based SEM needs hundreds
- Prediction focus: optimizes variance explained, making it ideal for predictive models
- Handles complexity: manages models with many latent variables, mediation chains, and non-recursive paths
- Non-normal data: does not require multivariate normality, robust to violations
- Practical software: widely implemented in accessible tools (SmartPLS, ADANCO, R packages)
- Model fit ambiguity: does not provide traditional goodness-of-fit indices (chi-square, CFI), making it harder to assess absolute model fit
- Assumption of reflective indicators: standard PLS assumes observed variables are reflections of latent constructs (not always true)
- Deliberate bias in composites: PLS composites are intentionally biased to predict outcomes, which can obscure construct validity
- Limited theory confirmation: not ideal for confirmatory hypothesis testing where precise fit matters
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
Should I use PLS-SEM or covariance-based SEM (CBSEM)?
Use PLS-SEM for prediction, exploratory research, small samples, or non-normal data. Use CBSEM for rigorous confirmatory testing, complex missing data, or when model fit indices matter. Increasingly, researchers use both: PLS to explore relationships, CBSEM to confirm findings.
What does R-squared mean in PLS-SEM?
R-squared is the proportion of variance in an endogenous variable explained by its predictors. It's the same as in OLS regression. In PLS-SEM, the goal is to maximize R-squared (prediction), not to achieve a pre-specified level as in CBSEM.
How do I assess measurement model validity in PLS-SEM?
Check: (1) indicator loadings (>0.70), (2) composite reliability (>0.70), (3) average variance extracted (AVE >0.50), and (4) discriminant validity (square root of AVE exceeds correlations with other constructs).
What are formative vs. reflective indicators?
Reflective indicators are caused by the latent construct (removing an indicator doesn't change the construct). Formative indicators cause the construct (removing an indicator changes the construct). PLS-SEM defaults to reflective; formative requires special estimation.
Can I use PLS-SEM with categorical outcome variables?
Standard PLS-SEM assumes continuous outcomes. For categorical outcomes, use PLS-PM (PLS for path modeling with discrete response) or adapt the approach using latent class analysis or mixture models.
Sources
- 1.Hair, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2017). A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM) (2nd ed.). Sage Publications.ISBN 9781483377445
- 2.Wold, H. (1985). Partial least squares. In S. Kotz & N. L. Johnson (Eds.), Encyclopedia of Statistical Sciences (Vol. 6, pp. 581-591). Wiley.ISBN 9780471822622
- 3.Chin, W. W. (2010). How to write up and report PLS analyses. In V. E. Vinzi, W. W. Chin, J. Henseler, & H. Wang (Eds.), Handbook of Partial Least Squares: Concepts, Methods and Applications (pp. 655-690). Springer.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 3). Partial Least Squares Structural Equation Modeling. ScholarGate. https://scholargate.app/psychometrics/pls-sem