NAVAL·SEM v0.9.0

Quantifies content validity by aggregating expert ratings at both the item and scale level, following the Polit & Beck (2006) framework.

• Item-level I-CVI computed as proportion of experts rating item as relevant (3 or 4).
• Scale-level S-CVI/Ave (average I-CVI) and S-CVI/UA (universal agreement) returned.
• Automatic pass/fail verdict against recommended thresholds (I-CVI ≥ 0.78, S-CVI/Ave ≥ 0.90).
• Accepts flexible expert matrix: any number of raters, any number of items.
• New POST /cvi API endpoint with full schema definitions.

Provides a full EFA workflow for scale development and pre-CFA exploration, supporting both oblique and orthogonal rotation strategies.

• Principal-axis factoring with promax (oblique) and varimax (orthogonal) rotation.
• Kaiser criterion and scree-plot eigenvalues returned for factor retention guidance.
• Factor loadings matrix, communalities, and percentage of variance explained per factor.
• Bartlett's test of sphericity and KMO measure of sampling adequacy included.
• Supports auto-extraction (eigenvalue > 1) or user-specified number of factors.
• New POST /efa API endpoint.

Assesses whether a construct behaves in theoretically expected ways by evaluating its bivariate correlations against directional hypotheses.

• Full inter-construct correlation matrix with Pearson r and p-values.
• Directional hypothesis testing: supply expected sign (+/−) per construct pair.
• Verdict returned per hypothesis (supported / not supported).
• Bootstrap-corrected confidence intervals around correlation estimates.
• Pydantic field fix: NomologicalResult.construct_name (alias construct) — silences BaseModel.construct() shadow warning without changing the JSON wire format.
• New POST /nomological API endpoint.

Full three-step MICOM (Measurement Invariance of Composites in PLS) procedure for assessing whether constructs are measured equivalently across groups.

• Step 1 — Configural invariance: verifies identical indicators and data treatment across groups.
• Step 2 — Compositional invariance: permutation test on correlation of composite scores across groups; p-value and 95% CI returned.
• Step 3 — Equality of means and variances: tests whether composite means and variances are equal across groups.
• Overall invariance verdict: full, partial, or not established.
• New POST /invariance API endpoint.

Implements the Effect Size Sensitivity Extension for Necessary Condition Analysis (Becker, Richter, Ringle & Sarstedt, 2026), stress-testing ceiling effect sizes by systematically removing threshold observations.

• Sensitivity sweep over CE-FDH and CR-FDH ceiling techniques — each observation in the ceiling band removed in turn and effect size recomputed.
• Joint-uniform benchmark: establishes expected effect-size distribution under no necessary condition.
• Permutation significance test for observed vs benchmark effect sizes.
• Returns per-threshold sensitivity profile: original effect size, range after removal, and robustness verdict.
• New POST /nca-esse API endpoint.
• New schemas: NCAESSEThresholdPoint, NCAESSEEntry, NCAESSEResult.
• Ref: J. Bus. Res. 206, 115920. doi:10.1016/j.jbusres.2025.115920

Five new modules and endpoints power the v0.9 analytical workflows:

Extended schemas.py and main.py cover all new analytical workflows with full Pydantic v2 model definitions and FastAPI route bindings.