VIF for formative blocks
Multicollinearity diagnostics per formative indicator. Standard quality criterion for Mode-B constructs with clear thresholds (< 5 uncritical, < 3 ideal).
Features
Everything you need for serious PLS-SEM.
From measurement model to publication-ready report, organised by use case. Validated against established reference values from the PLS-SEM literature.
Measurement model
Reliability and validity of your constructs. Reflective (Mode A) and formative (Mode B) in one editor.
Loadings, weights & cross-loadings
Full outer-model panel per indicator. Loadings for reflective blocks, weights for formative blocks, plus common-method diagnostics via cross-loadings.
AVE > 0.5 convergent validity
Reliability: Cronbach, ρ_A, DG ρ, AVE
Four reliability measures in one table. Cronbach as the lower bound, Dijkstra-Henseler ρ_A as the point estimate, DG ρ as the upper bound, AVE as the convergence measure.
Unidimensionality
Eigenvalue check per reflective block (Dillon-Goldstein, first and second eigenvalue). Instantly visible when a block measures more than one dimension.
Discriminant validity
Three independent perspectives on construct distinctness in one panel.
HTMT
Heterotrait-monotrait ratio after Henseler, Ringle & Sarstedt (2015). Standard in every PLS-SEM paper since 2015. Threshold 0.85 (strict) or 0.90 (liberal).
HTMT2
Geometric-mean refinement after Roemer, Schuberth & Henseler (2021). Unbiased under unequal loadings within a block, more conservative than classic HTMT.
Fornell-Larcker criterion
Compares √AVE of each LV with its maximum correlation to other LVs. Classic test since 1981, still required by many journals.
Structural model
Path coefficients, effect sizes, and effect decomposition. Direct, indirect, specific.
Path coefficients & inner VIF
Standardised beta values for every structural path, plus inner VIF for diagnosing multicollinearity between predictor LVs.
R², adjusted R² & BIC
Explanatory power per endogenous LV plus a sample-size-adjusted variant and Bayesian information criterion for model comparison.
0.02 · 0.15 · 0.35 Cohen f² thresholds
f² effect sizes
Cohen’s f² per path: measures the unique contribution of each predictor to the R² of the endogenous LV. 0.02 / 0.15 / 0.35 as rule of thumb.
Specific indirect effects
Every mediation chain listed separately, with point estimate and bootstrap CI per chain. Not just the total indirect effect, but each individual path’s contribution.
Q² > 0 predictive relevance
Q² predictive relevance
Stone-Geisser Q² via blindfolding. Identifies whether the model has predictive power beyond the in-sample fit.
Inference & validation
Uncertainty quantification and sample comparison. Standardised procedures from the PLS-SEM community.
Bootstrap
Resampling with replacement for t-values, p-values and confidence intervals. Runs on Cloud Run, 5000 iterations typically in 2 to 5 minutes.
Moderation (two-stage)
Interaction terms in the structural model, two-stage approach after Henseler & Chin (2010). One click in the editor.
Multi-group analysis (MGA)
Tests for structural differences between groups. Bootstrap-based (Henseler MGA) and permutation-based in the same view.
SRMR < 0.08 · GoF ≥ 0.36 approximate fit indices
Model fit (SRMR, d_ULS, GoF)
Approximate model fit indices: SRMR (< 0.08 good), d_ULS for saturated/estimated, goodness-of-fit after Tenenhaus et al. (2005).
Advanced methods
Modern PLS-SEM techniques that are often missing from standard tools. One editor, one click.
FIMIX-PLS
Finite-mixture segmentation reveals unobserved subgroups. Surfaces heterogeneous effects before they bias your results.
PLSpredict & IPMA
k-fold cross-validation with the full Shmueli panel (RMSE, MAE, MAPE versus LM) and importance-performance map for management prioritisation.
PLSc (consistent PLS)
Dijkstra-Henseler bias correction for reflective measurement models. Delivers ρ_A per LV, disattenuated correlations and corrected paths.
Higher-order constructs
Disjoint two-stage workflow for hierarchical models. All four canonical types (R-R, R-F, F-R, F-F) plus nested HOCs.
Gaussian copula
Park-Gupta / Hult-et-al. endogeneity test for structural predictors. Detects bias without instrumental variables, with admissibility check.
Newton & PCA schemes
Quasi-Newton inner-weighting for more stable convergence, Lohmöller’s PCA scheme as an alternative to centroid and path.
Reporting & export
From model directly to paper. Open formats, no lock-in.
PDF publication-ready
Publication PDF report
Publication-ready PDF report with all tables, structural diagram and methods section. Drop straight into your appendix.
XLSX & LaTeX tables
Every table as an XLSX sheet and as a LaTeX snippet, one click to copy. Saves hours of reformatting in Overleaf.
Report JSON bundle
The full report as structured JSON for reproducibility statements and machine-readable analysis. Format documented, no proprietary container.
SmartPLS import (.splsm)
Import existing SmartPLS models without rebuilding. One-step migration with no data loss.
Docker · GPL · On-Premise cloud or self-hosted
Self-host & teams
Docker image for sensitive data, multi-user projects with version comparison for research groups. Cloud or on-premise, your choice.
Ready to run your first model?
Free account, no credit card required. The first projects run in the cloud, one click to self-host when you’re ready.