Added
Moderation Analysis
engine_moderation.py · product-of-composites
A dedicated moderation engine brings first-class interaction modelling to NAVAL·SEM.
- Product-of-composites estimation for latent moderation.
- Automatic interaction-term detection using lavaan-style
X*Msyntax — no extra setup required. - Bootstrap confidence intervals for all interaction effects.
- Simple-slope analysis at low (−1 SD), mean, and high (+1 SD) moderator values.
- ΔR² and Cohen's f effect-size reporting per moderation effect.
- Manifest-variable OLS fallback for models without latent constructs.
Importance-Performance Map Analysis (IPMA)
engine_ipma.py · construct prioritisation
The new IPMA engine surfaces which constructs deserve management attention by combining path-coefficient importance with rescaled performance.
- Total-effect importance computation for all predictor constructs.
- 0–100 performance rescaling of latent variable scores.
- Target-construct analysis endpoint (
POST /ipma) and dedicated reporting view.
Necessary Condition Analysis (NCA)
engine_utils.py · ceiling techniques
NCA identifies the necessary (but not sufficient) conditions in your model — constructs whose absence guarantees a low outcome, regardless of other predictors.
- CE-FDH and CR-FDH ceiling-line techniques, automatically selected per variable pair.
- Permutation-based significance testing for each ceiling effect.
- Effect-size classification (small / medium / large) and bottleneck analysis support.
- NCA result schemas and structured reporting objects.
Conditional Process Analysis
engine_mod_mediation.py · Hayes PROCESS compatibility
Moderated mediation is now fully supported — estimate how and when indirect effects operate, following the Hayes PROCESS framework.
- Conditional indirect-effect estimation at low, mean, and high moderator values.
- Index of Moderated Mediation (IMM) with bootstrap confidence intervals.
- Support for Hayes PROCESS-style Models 7, 14, and 58/59 out of the box.
- Bootstrap CIs for all conditional indirect effects, fully integrated with the existing bootstrap engine.
New Backend Engines
app/ · API routes & schemas
Four new modules power the v0.7 analytical workflows:
app/engine_moderation.py
app/engine_mod_mediation.py
app/engine_ipma.py
app/engine_utils.py
Extended API routes and schema definitions cover all new analytical workflows, including new POST /moderation, POST /ipma, POST /nca, and POST /conditional-process endpoints.
Improved
- Model Summary object. A new
ModelSummaryreporting object consolidates the most important fit indices, validity metrics, and path estimates into a single high-level digest — designed for quick UI panels and export. - Reproducibility fingerprint. Every completed run now generates a SHA-256 audit record covering the model syntax hash, data hash (sorted column hashes), algorithm choice, environment versions, and key fit metrics — retrievable via
GET /fingerprint/{id}. - Automatic reverse scoring. Variables following the
_Rnaming convention are now detected and reverse-scored automatically before model fitting — no manual pre-processing step needed. - Parser handling. Improved syntax parsing for advanced moderation and conditional-process expressions, including nested interaction terms and multi-moderator specifications.
- SEM reporting & result aggregation. Structural and measurement results are now aggregated more consistently across CB-SEM and PLS-SEM paths, reducing edge-case discrepancies in multi-group and HOC workflows.
- API documentation. Expanded endpoint coverage across all new analytical routes, with updated schema definitions and example payloads in the docs.
Changed
- Refactored into dedicated engines. v0.7 functionality previously in
engine_v07.pyhas been split into the four specialist modules above. The legacy file has been removed. - Frontend updated. The static frontend (
static/index.html) has been updated to surface new workflows — Moderation, IPMA, NCA, and Conditional Process each have dedicated UI panels and results sections. - Version bumped to 0.7.1.