Embedded Files
In SmartPLS 4, robustness check is not a single button. It is done through a set of additional analyses to verify whether your main results remain stable when you examine issues such as nonlinearity, measurement invariance, observed and unobserved heterogeneity, endogeneity, and predictive performance. The Primer places these topics under advanced methods, while Advanced Issues discusses them in more detail, especially nonlinear relationships, MICOM and MGA, and FIMIX-PLS / PLS-POS.
Practical wayS to do robustness check in SmartPLS 4
Practical wayS to do robustness check in SmartPLS 4
1. Run the main PLS-SEM model first
1. Run the main PLS-SEM model first
Complete the standard analysis first:
measurement model assessment
structural model assessment
bootstrapping
R², f², Q², and if relevant, PLSpredict
This gives you the baseline model. Robustness checks are then used to see whether the main findings remain stable. The Primer clearly separates the standard model evaluation from advanced methods.
2. Check for nonlinearity
2. Check for nonlinearity
The Advanced Issues book includes a dedicated section on Nonlinear Relationships, including quadratic effects modeling, evaluation, and interpretation. This means one important robustness check is to test whether a relationship you assumed to be linear is actually nonlinear.
In practice:
add a quadratic or nonlinear term for the path you suspect may be nonlinear,
re-estimate the model,
assess whether the nonlinear effect is significant.
Interpretation:
if the nonlinear effect is not significant, the original linear result is more robust;
if it is significant, the earlier linear specification may be incomplete.
3. Check observed heterogeneity using MICOM and MGA
3. Check observed heterogeneity using MICOM and MGA
The Advanced Issues book outlines this procedure clearly:
Step 1: Configural invariance
Step 2: Compositional invariance
Step 3: Equality of composite mean values and variances
followed by Multigroup Analysis, including PLS-MGA and the Permutation Test.
In practice:
Split the sample into groups, for example male vs female, Indonesia vs Malaysia, public vs private.
Run MICOM.
If at least partial invariance is established, continue with MGA.
Compare the path coefficients across groups.
Interpretation:
if the path relationships are similar across groups, your findings are more robust;
if they differ significantly, the results are context-specific rather than universal.
The Primer also treats measurement model invariance and multigroup analysis as advanced methods that are necessary when comparing groups.
4. Check unobserved heterogeneity using FIMIX-PLS and PLS-POS
4. Check unobserved heterogeneity using FIMIX-PLS and PLS-POS
The Advanced Issues book provides a step-by-step procedure:
Step 1: Run the FIMIX-PLS Procedure
Step 2: Determine the Number of Segments
Step 3: Run the PLS-POS Procedure
Step 4: Explain the Latent Segment Structure
Step 5: Estimate Segment-Specific Models.
In practice:
Run FIMIX-PLS to see whether hidden segments may exist in the data.
Identify the most plausible number of segments.
Run PLS-POS.
Estimate and compare the model for each segment.
Interpretation:
if the data essentially support a single-segment solution, the pooled results are more robust;
if hidden segments exist and path coefficients vary across them, the overall model may hide important subgroup differences.
The Primer also lists treating observed and unobserved heterogeneity and uncovering unobserved heterogeneityamong the advanced PLS-SEM topics.
5. Check endogeneity
5. Check endogeneity
The Primer explicitly includes examining endogeneity among the advanced methods. This is important especially for explanatory models, because endogeneity can bias path estimates and lead to incorrect conclusions.
Conceptually, the procedure is:
identify predictors that may be endogenous,
apply an appropriate endogeneity assessment,
examine whether the key structural relationships remain stable.
The key robustness question is whether your main path coefficients remain trustworthy after potential endogeneity is considered.
6. Check predictive robustness using PLSpredict
6. Check predictive robustness using PLSpredict
The Primer includes the model’s predictive power as part of structural model evaluation and emphasizes PLSpredict as an important modern development in PLS-SEM.
In practice:
run PLSpredict
inspect Q²_predict
compare the prediction errors of the PLS model with the benchmark linear model
Interpretation:
if most indicators have Q²_predict > 0 and the PLS model performs better, or at least competitively, compared to the linear benchmark, this supports predictive robustness;
if many indicators have Q²_predict < 0, the model is weak from a prediction perspective.
Recommended robustness workflow
Recommended robustness workflow
A practical robustness workflow in SmartPLS 4 would be:
Estimate the main model
Check nonlinearity
Run MICOM and MGA if group comparisons are relevant
Run FIMIX-PLS / PLS-POS if unobserved heterogeneity is plausible
Assess endogeneity if the model is explanatory
Run PLSpredict for predictive robustness
Conclusion
Conclusion
In SmartPLS 4, robustness check is carried out through a set of additional analyses rather than through one single menu. Based on Hair et al., the most important robustness checks involve nonlinear relationships, measurement invariance and multigroup differences, unobserved heterogeneity, endogeneity, and predictive performance. If the substantive findings remain consistent across these checks, the results can be considered more robust.
Further Readings
Further Readings
Sarstedt, M., Ringle, C. M., Cheah, J. H., Ting, H., Moisescu, O. I., & Radomir, L. (2020). Structural model robustness checks in PLS-SEM. Tourism Economics, 26(4), 531-554.
Vaithilingam, S., Ong, C. S., Moisescu, O. I., & Nair, M. S. (2024). Robustness checks in PLS-SEM: A review of recent practices and recommendations for future applications in business research. Journal of Business Research, 173, 114465.
Google Sites
Report abuse