FIMIX-PLS (Finite Mixture PLS) detects unobserved heterogeneity in your PLS-SEM model by identifying latent segments with distinct path coefficient profiles.
- Expectation-Maximisation (EM) algorithm with configurable segment count (k).
- Segment-specific path coefficients and R² values for every endogenous construct.
- Information criteria (AIC, BIC, AIC3, CAIC, EN) for optimal segment selection.
- Relative entropy reporting to assess segment separation quality.
- Posterior segment membership probabilities for each observation.
- New
POST /fimixAPI endpoint with full schema definitions.
PLS-POS complements FIMIX-PLS with a distance-based segmentation approach, assigning observations to segments by minimising within-segment residual heterogeneity.
- Iterative residual-based segment assignment without distributional assumptions.
- Segment solutions for k = 2 to 5 with automatic convergence detection.
- Per-segment path coefficients, loadings, and predictive accuracy metrics.
- Segment size and composition reporting for downstream interpretation.
- New
POST /posAPI endpoint.
The Gaussian Copula approach (Park & Gupta, 2012) corrects for endogeneity caused by non-normal, non-elliptical predictor distributions — a common threat to PLS-SEM validity in survey and secondary data.
- Automatic non-normality screening using Henze-Zirkler and skewness tests.
- Copula term construction and augmented regression for each flagged predictor.
- Bootstrap confidence intervals for copula coefficients.
- Corrected path coefficients alongside original estimates for comparison.
- Cohen's f² effect size for the endogeneity correction increment.
- New
POST /gaussian-copulaAPI endpoint.
Generate a complete, publication-ready PDF report from any completed SEM run — covering measurement model, structural model, fit indices, and all advanced analysis results.
- Single-click export from the results panel to a self-contained PDF.
- Sections for measurement validity (AVE, CR, HTMT), structural paths, bootstrapped CIs, and model fit.
- Advanced results (FIMIX, PLS-POS, Gaussian Copula, IPMA, NCA) included when run.
- APA-style table formatting throughout.
- New
POST /export/pdfAPI endpoint returning a binary PDF stream.
Four new modules power the v0.8 analytical workflows:
Extended API routes and schema definitions cover all new analytical workflows, including new POST /fimix, POST /pos, POST /gaussian-copula, and POST /export/pdf endpoints.
- Centralised versioning.
app/version.pynow holds the single source of truth forAPP_VERSION; CI injects the correct version from the git tag at build time, eliminating all hardcoded version strings across the codebase. - CI/CD pipeline. The GitHub Actions release workflow has been simplified — the sync-version-back-to-master step has been removed, resolving recurring branch-protection failures. All three platform builds (Windows EXE + MSI, macOS DMG, Linux binary + .deb) now complete cleanly.
- Results aggregation. The model summary object now consolidates FIMIX, PLS-POS, and Gaussian Copula outputs alongside existing CB-SEM / PLS-SEM results for unified reporting.
- Frontend panels. New UI sections for FIMIX-PLS, PLS-POS, Gaussian Copula, and PDF export are integrated into the results tab with consistent layout and colour coding.
- Version bumped to 0.8.0. This is the first release with CI-injected versioning — the
_DEFAULT_APP_VERSIONinapp/version.pyis set to"dev"on master and overwritten at build time from the git tag. - Distribution. All five release artifacts (EXE, MSI, DMG, Linux binary, .deb) are now consistently published to GitHub Releases and mirrored on SourceForge.