GSCA in SmartPLS: A Component-Based Approach Distinct from PLS-SEM

GSCA (Generalized Structured Component Analysis) is a component-based algorithm integrated into SmartPLS 4. It estimates the entire model using alternating least squares and provides global fit indices unavailable in classical PLS-SEM under the same logic. It is a separate method rather than an improved PLS-SEM variant.

PLS-SEM & SMARTPLSPrepared by Lê Minh6 min

Start reading

Composite assessment needs appropriate evidence

Concept
Clarify the composite’s role
Specification
Justify how it is formed
Assessment
Examine the relevant evidence
Limitations
State the scope of the approach

Conceptual diagram. Do not automatically interchange common-factor and composite criteria.

Source: Lê Minh · Conceptual illustration, not software output.

1. What Is GSCA?

Generalized Structured Component Analysis (GSCA), proposed by Hwang and Takane (2004), is a component-based SEM method. Like PLS-SEM, it constructs composites as weighted combinations of observed variables and estimates structural relationships between those composites rather than between latent variables in the classical CB-SEM sense.

The difference lies in how weights and path coefficients are estimated. Hwang and Takane (2004) built GSCA around a single least squares criterion covering both measurement and structural models, optimizing it jointly. The authors used this approach to address a limitation recognized in PLS-SEM at the time: classical PLS-SEM does not optimize a single global criterion and therefore lacks an overall fit index in the sense familiar from CB-SEM.

Composites in GSCA

A composite is a weighted linear combination of its observed variables. GSCA supports composites in reflective and formative measurement models and more complex structures such as mediation.

2. Alternating Least Squares Estimation

GSCA estimates parameters with an alternating least squares (ALS) algorithm. It iteratively updates indicator weights to maximize the variance in indicators and dependent composites explained by the model, simultaneously minimizing unexplained variance throughout the model.

Traditional PLS-SEM instead estimates composite scores through a sequence of iterative simple and multiple regressions by variable block, converging without optimizing a single overall criterion. Because GSCA jointly optimizes a global criterion, it provides measures of fit for the entire model in a way not available in classical PLS-SEM.

3. Four Steps in GSCA Analysis

Specify the model

Define the composites, their indicators, measurement model types (reflective or formative), and structural relationships according to the research hypotheses.

Estimate the weights

Run ALS to jointly estimate indicator weights and path coefficients, minimizing unexplained variance throughout the model.

Assess the model

Examine global fit indices (FIT, AFIT, and GFI) alongside bootstrap results to test the significance of path coefficients and indicator weights.

Interpret the results

Interpret weights, path coefficients, and model fit in relation to the research question, similarly to PLS-SEM interpretation but using GSCA-specific indices.

4. SmartPLS Settings to Understand

When running GSCA in SmartPLS 4, two groups of settings deserve attention: starting weights and stopping conditions.

Table 1
Setting	Option	Meaning
Starting weights	Standard (default)	All starting weights are set to +1.
Starting weights	Individual	Users assign an initial weight to each indicator, for example +1 for a key indicator and 0 for the others.
Maximum iterations	Fixed at 3.000 iterations	Cannot be adjusted in SmartPLS 4.
Stopping criterion	Fixed at 10^-7	Cannot be adjusted in SmartPLS 4.

Significance testing for GSCA weights and path coefficients uses SmartPLS's general bootstrap procedure. The number of bootstrap samples and confidence interval type are therefore configured in the same way as for PLS-SEM.

5. Global Fit Indices: FIT, AFIT, and GFI

Because estimation follows a global criterion, GSCA provides fit indices for the entire model under a logic different from traditional PLS-SEM. Three common indices are FIT, AFIT (FIT adjusted for the number of parameters), and GFI (Goodness of Fit Index).

Avoid choosing cutoffs intuitively

Cho, Hwang, Sarstedt, and Ringle (2020) devoted a study to appropriate cutoffs for GSCA's FIT, AFIT, and GFI because these indices do not share the scale or interpretation of familiar CB-SEM indices such as CFI and RMSEA. Transferring cutoff intuitions from CB-SEM or PLS-SEM to them is unsupported.

A FIT or GFI value that sounds high does not automatically establish good fit under GSCA standards. Consult cutoff research specifically for GSCA instead of borrowing another method's conventions.

6. When to Consider GSCA instead of PLS-SEM

Table 2
Aspect	GSCA	PLS-SEM
Estimation logic	Joint optimization of a global least squares criterion through ALS.	Composite score estimation through iterative blockwise regressions, converging without a single global criterion.
Overall fit indices	Available (FIT, AFIT, and GFI), requiring specific cutoffs from Cho, Hwang, Sarstedt, and Ringle (2020).	No overall fit index under the same logic; model assessment uses separate criteria such as reliability, AVE, HTMT, R², and Q².
Supported measurement models	Reflective and formative models, with more complex structures such as mediation.	Reflective and formative models, with extensions such as higher-order constructs and moderation.
Prevalence in theses and field-specific journals	Newer and less common, with fewer applied studies than PLS-SEM.	Widely used in the social sciences and business management, with extensive reporting guidance.

GSCA deserves consideration when the research requires a global fit index or seeks to use ALS's global optimization properties. PLS-SEM has more reporting guidance and publication precedents when the objective is comparison with most studies in a field or adherence to widely standardized reporting procedures.

7. Common Mistakes

Table 3
Weak wording	Why it is inappropriate	Recommended wording
"GSCA is an improved version of PLS-SEM."	The methods have different estimation logics and optimization criteria; they are not earlier and later versions of the same algorithm.	Describe GSCA as a separate component-based SEM method developed by Hwang and Takane (2004).
"Use CFI > 0,90 or PLS-SEM standards to assess GSCA's FIT."	FIT, AFIT, and GFI do not share the scale and interpretation of CB-SEM indices or PLS-SEM assessment criteria.	Refer to GSCA-specific cutoffs from Cho, Hwang, Sarstedt, and Ringle (2020).
"Choose GSCA because the software has a button for it; no further reason is needed."	Selecting an estimation method requires justification based on the research question and model characteristics, beyond software availability.	Explain why GSCA was chosen, such as a need for global fit indices or ALS optimization properties.
"GSCA and PLS-SEM always give identical results, so either can be used."	The algorithms have different estimation mechanisms and need not produce identical results, particularly with small samples or complex models.	Report the method and rationale clearly, avoiding an assumption of equivalent results.

8. Checklist before Reporting

Have you justified choosing GSCA over PLS-SEM or CB-SEM based on the research question?
Have you described each composite's reflective or formative measurement model?
Have you stated whether Standard or Individual starting weights were used?
Have you reported FIT, AFIT, and GFI with reference cutoffs from Cho, Hwang, Sarstedt, and Ringle (2020)?
Have you used bootstrapping to test weights and path coefficients?
Have you avoided describing GSCA as a PLS-SEM upgrade in the methodology?

Conclusion

GSCA offers component-based estimation with a mathematical foundation distinct from PLS-SEM: joint optimization of a global criterion through alternating least squares rather than convergence through blockwise regressions. Consequently, GSCA provides global fit indices unavailable in classical PLS-SEM under the same logic. Those indices require their own cutoffs, which cannot be borrowed from CB-SEM or PLS-SEM.

Choosing GSCA should follow from the research question and model characteristics, beyond its integration in SmartPLS 4. Given fewer publication precedents than PLS-SEM, GSCA users should also be prepared to explain the method to unfamiliar readers.

References

Hwang, H., & Takane, Y. (2004). Generalized structured component analysis. Psychometrika, 69(1), 81-99. https://doi.org/10.1007/BF02295841
Hwang, H., & Takane, Y. (2014). Generalized structured component analysis: A component-based approach to structural equation modeling. Chapman & Hall/CRC. https://doi.org/10.1201/b17872
Cho, G., Hwang, H., Sarstedt, M., & Ringle, C. M. (2020). Cutoff criteria for overall model fit indexes in generalized structured component analysis. Journal of Marketing Analytics, 8, 189-202. https://doi.org/10.1057/s41270-020-00089-1
Cho, G., & Hwang, H. (2024). Generalized structured component analysis accommodating convex components: A knowledge-based multivariate method with interpretable composite indexes. Psychometrika, 89(1), 241-266. https://doi.org/10.1007/s11336-023-09944-3

Academic note: This article presents GSCA's general principles based on Hwang and Takane's original work and SmartPLS guidance. Official SmartPLS documentation does not directly compare GSCA with PLS-SEM or recommend exactly when to choose each. The comparisons and selection recommendations here are reasoned inferences from their technical characteristics rather than quotations from a single official recommendation.