Quick answer: Use PLS-SEM if your goal is prediction or exploring a theory that is still developing, if your model has formative constructs or many paths, or if your sample is modest. Use CB-SEM if you are confirming a well-established theory, your data is close to normal, and your examiners expect global model-fit indices such as CFI and RMSEA. The decision should follow from your research objective, not from which software you happen to know.
Many scholars choose their analysis method only after the data is collected, and then spend weeks defending that choice to their committee. If you are planning your phd management thesis, the pls sem vs cb sem decision belongs at the design stage, ideally in your project synopsis, where you justify the method before you collect a single response.
What Is PLS-SEM vs CB-SEM? Key Differences Explained
Both are forms of structural equation modelling (SEM). Each tests a measurement model (how survey items reflect constructs) and a structural model (how constructs influence one another) at the same time. They estimate these models in fundamentally different ways.
- CB-SEM (covariance-based SEM) estimates the model parameters so that the covariance matrix implied by the model is as close as possible to the observed covariance matrix. It usually relies on maximum likelihood estimation and treats constructs as common factors. Common software includes AMOS, LISREL, Mplus and R's lavaan.
- PLS-SEM (partial least squares SEM) is variance-based and composite-based. It aims to maximise the explained variance (R²) of the dependent constructs. Most management researchers run it in SmartPLS.
In short, CB-SEM asks whether your theory fits the data, while PLS-SEM asks how well your model explains and predicts outcomes. This distinction is central to any pls sem vs cb sem for phd thesis discussion, and Hair, Risher, Sarstedt and Ringle make it explicitly in their widely cited 2019 guidance in the European Business Review.
When to Use PLS-SEM in Your PhD Management Thesis
Knowing when to use pls sem vs cb sem begins with your research question. PLS-SEM is usually the stronger choice when:
- Your goal is prediction or theory development. Examples include studies on emerging topics such as AI adoption, gig-work engagement or digital banking trust, where theory is still evolving.
- Your model includes formative constructs. These are indices built from their indicators, such as "service quality" measured through distinct dimensions.
- Your model is complex. Many constructs, mediators, moderators or higher-order constructs are handled well in PLS-SEM.
- Your sample is moderate. PLS-SEM converges with smaller samples, although that does not make a small sample adequate.
- Your data is non-normal. Bootstrapping in PLS-SEM does not assume a normal distribution.
For reporting, examiners now expect more than path coefficients. Plan to report outer loadings, composite reliability, AVE, discriminant validity through HTMT (Henseler, Ringle & Sarstedt, 2015), R², f², and out-of-sample predictive power through PLSpredict (Shmueli et al., 2019).
Struggling to set up your model? Our team supports the entire analysis chapter of a PhD management thesis, including model specification, SmartPLS output and interpretation. ? Explore PhD Management Thesis Support
When CB-SEM Is the Better Choice for Theory Testing
CB-SEM remains the gold standard for confirming established theories. Choose it when:
- You are testing well-established frameworks such as TAM, UTAUT or the Theory of Planned Behaviour in a new context.
- All your constructs are reflective, meaning the items are interchangeable expressions of one underlying factor.
- You need global model-fit statistics to support your claims.
- Your sample is large, typically 200 or more cases, with approximately normal data.
- You plan to run confirmatory factor analysis (CFA) and test measurement invariance across groups.
This is also the core of the smartpls vs amos for phd research debate. AMOS (or lavaan) suits confirmatory, fit-driven studies, while SmartPLS suits predictive, complex models. Some scholars use both: they run a CFA in AMOS to validate scales and then use PLS-SEM for prediction. If you do this, justify it clearly in your methodology chapter.
Sample Size, Data Normality & Model Fit: Quick Comparison
| Primary goal | Prediction, explanation | Theory confirmation |
| Sample size | Works with smaller samples; use the inverse square root method (Kock & Hadaya, 2018) or a power analysis | Usually 200+; larger for complex models |
| Data distribution | No normality assumption (bootstrapping) | ML assumes multivariate normality; robust estimators are available |
| Construct type | Reflective and formative | Mainly reflective |
| Model fit | Limited; SRMR is commonly reported (< 0.08) | Chi-square, CFI/TLI (≥ 0.90–0.95), RMSEA (≤ 0.06–0.08), SRMR (≤ 0.08) |
| Typical software | SmartPLS | AMOS, LISREL, lavaan, Mplus |
Do not rely on the old "10-times rule" for PLS-SEM sample size. Methodologists have criticised it as unreliable, and reviewers increasingly reject it. A G*Power analysis or the inverse square root method is far easier to defend.
Need help with sample size justification or a defensible methods section for your project synopsis? ? Talk to an Anushram Research Expert Today
PLS-SEM or CB-SEM in 2026: Which Do Reviewers Prefer?
Neither method is universally preferred. What reviewers and examiners value most in 2026 is justification. PLS-SEM has faced methodological criticism (notably from Rönkkö and colleagues) when it is used without a clear reason, and in response, journals and doctoral committees now expect you to explain why your method fits your objective.
Here is what consistently passes review:
- Match the method to the aim. Use prediction for PLS-SEM and confirmation for CB-SEM.
- Cite methodological authorities. Examples include Hair et al. (2019, 2022) for PLS-SEM, and Kline's Principles and Practice of Structural Equation Modeling or Hu & Bentler (1999) for CB-SEM fit criteria.
- Report transparently. Include bootstrapping settings, HTMT, and either PLSpredict or full fit indices.
- Consider consistent PLS (PLSc) if you use PLS-SEM with purely reflective constructs.
So, which sem method is best for management research? The best method is the one aligned with your research question and defended with current literature. This principle of methodological rigour runs through Indian scholarship, from the thesis of dr br ambedkar on provincial finance at Columbia to today's doctoral work. Requirements also differ by university. For example, health-management scholars preparing an rguhs thesis submission should check their university's statistical reporting norms before finalising the analysis chapter.
Want a reviewer-ready analysis chapter? ? Get Your Free PhD Thesis Consultation Now
Decision Checklist: PLS-SEM vs CB-SEM for PhD Thesis
- Prediction, complex model, formative constructs → PLS-SEM (SmartPLS)
- Theory confirmation, reflective constructs, large normal sample → CB-SEM (AMOS/lavaan)
- Unsure? Fix the decision in your project synopsis and get expert review before collecting data.
FAQs
Q1. Which SEM method is best for management research?
A. Neither is best in all cases. PLS-SEM suits predictive and exploratory research, and CB-SEM suits confirmatory theory testing. Choose based on your research objective.
A. Neither is best in all cases. PLS-SEM suits predictive and exploratory research, and CB-SEM suits confirmatory theory testing. Choose based on your research objective.
Q2. When should I use PLS-SEM vs CB-SEM in a PhD thesis?
A. Use PLS-SEM for prediction, formative constructs, complex models or non-normal data. Use CB-SEM when testing established theory with a large sample and when global fit indices are required.
A. Use PLS-SEM for prediction, formative constructs, complex models or non-normal data. Use CB-SEM when testing established theory with a large sample and when global fit indices are required.
Q3. SmartPLS vs AMOS for PhD research: which is easier?
A. SmartPLS has a simpler interface for complex and predictive models. AMOS is the standard for CFA and fit-based confirmatory analysis. Your research aim should decide, not convenience.
A. SmartPLS has a simpler interface for complex and predictive models. AMOS is the standard for CFA and fit-based confirmatory analysis. Your research aim should decide, not convenience.
Q4. Is a small sample acceptable for PLS-SEM?
A. PLS-SEM can run on smaller samples, but you still need adequate statistical power. Justify your sample size with a power analysis or the inverse square root method.
A. PLS-SEM can run on smaller samples, but you still need adequate statistical power. Justify your sample size with a power analysis or the inverse square root method.
Q5. Can I use both PLS-SEM and CB-SEM in one PhD management thesis?
A. Yes, as long as each method serves a clearly stated purpose and the combination is justified in your methodology chapter.
A. Yes, as long as each method serves a clearly stated purpose and the combination is justified in your methodology chapter.
? Don't Let the Wrong SEM Method Delay Your PhD
Anushram's research experts help management scholars choose, run and defend the right SEM approach, from synopsis to final submission. Sign Up for Expert PhD Thesis Support →
