Statistical Validity Score

The Validity Score is Licklider's pre-export quality gate. It evaluates every figure across eight dimensions of analytical quality and produces a composite score. A figure cannot be exported in publication-ready mode unless all blocking dimensions pass.

What is the Validity Score?

Each of the eight dimensions corresponds to a category of reporting error that is both common in published research and detectable from the analytical record. The score does not penalize low statistical power or borderline p-values — it penalizes missing disclosures, undeclared decisions, and configuration errors that prevent a reader from evaluating the result.

Score interpretation: Each dimension is scored Pass / Warning / Fail. A single Fail in a blocking dimension prevents publication-ready export. Warnings require acknowledgment. A figure that passes all eight dimensions receives the Validity Score badge in its export package.

Eight Dimensions

4.1.1 Normality & Homoscedasticity

Shapiro-Wilk / KS / D'Agostino normality tests + Levene homoscedasticity — auxiliary diagnostics for total validity assessment.

4.1.2 Effect Size Reporting

Cohen's d / η² / r auto-calculated by test type and required in all inferential exports.

4.1.3 CI Reporting

95% CI auto-calculated for all analyses. Null CI cases disclosed explicitly.

4.1.4 Power & Sample Size

G*Power-equivalent calculation generates justification text from declared effect size and α.

4.1.5 Multiple Comparison Compliance

Correction selection forced for ≥3 group pairwise comparisons. Uncorrected publication output blocked.

4.1.6 Replication Type Declaration

Independent vs. technical replication declared and verified against subject ID column structure.

4.1.7 Missing Data Disclosure

Rate, MCAR/MAR/MNAR mechanism, LOD pattern detection — all required before export.

4.1.8 Outlier Criteria Pre-Registration

Criterion declared before exclusion. Raw / excluded / robust estimates shown in parallel.