Description
Submission Confidence Index Engine.
Description
Converts standardized R4SUB (R for Regulatory Submission) evidence into indicator scores, pillar scores, and a Submission Confidence Index (SCI). Provides sensitivity analysis, explainability tables, and decision band classification to answer the question: are we ready for regulatory submission.
README.md
r4subscore
r4subscore is the scoring and calibration engine of the R4SUB ecosystem.
It converts standardized evidence (from r4subcore and companion packages like r4subtrace) into:
- Indicator Scores (per-indicator 0--1)
- Pillar Scores (per-domain 0--1)
- Submission Confidence Index (SCI, 0--100)
- Decision Bands (Ready / Minor Gaps / Conditional / High Risk)
- Sensitivity Analysis (SCI stability under weight variations)
- Explainability Tables (which indicators drive SCI up/down)
It answers the executive question:
Are we ready for regulatory submission -- and how confident are we?
Installation
pak::pak(c("R4SUB/r4subcore", "R4SUB/r4subscore"))
Quick Start
library(r4subcore)
library(r4subscore)
# assume ev is a validated evidence table
pillar_scores <- compute_pillar_scores(ev)
sci <- compute_sci(pillar_scores)
sci$SCI
sci$band
Core Functions
| Function | Purpose |
|---|---|
sci_config_default() | Pillar weights + decision bands config |
classify_band() | Classify an SCI value into a decision band |
compute_indicator_scores() | Severity-weighted indicator-level scores |
compute_pillar_scores() | Aggregate indicators into pillar scores |
compute_sci() | Compute SCI (0--100) + band classification |
sci_sensitivity_analysis() | SCI under alternative weight scenarios |
sci_explain() | Top loss contributors + pillar breakdown |
SCI Bands
| SCI | Band | Interpretation |
|---|---|---|
| 85--100 | ready | Ready for Submission |
| 70--84 | minor_gaps | Minor Gaps to Address |
| 50--69 | conditional | Conditional -- Address Key Issues |
| 0--49 | high_risk | High Risk |
Scoring Logic
- Each evidence row gets a weighted score:
result_score * (1 - severity_weight) - Indicator scores = mean weighted score per indicator
- Pillar scores = mean indicator score per domain
- SCI = weighted sum of pillar scores x 100
License
MIT.