r4subscore is the scoring and calibration engine of the R4SUB ecosystem. It converts standardized evidence (from r4subcore and companion packages) into a Submission Confidence Index (SCI) — a single 0–100 score with decision bands, explainability tables, and sensitivity analysis.
Are we ready for regulatory submission — and how confident are we?
Quick Start
library(r4subcore)
library(r4subscore)
pillar_scores <- compute_pillar_scores(ev)
sci <- compute_sci(pillar_scores)
sci$SCI # 0–100
sci$band # "ready", "minor_gaps", "conditional", or "high_risk"SCI Decision 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 (quality, trace, risk, usability)
- SCI = weighted sum of pillar scores × 100
Key Functions
| Function | Purpose |
|---|---|
sci_config_default() |
Pillar weights and decision bands configuration |
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) and band classification |
sci_sensitivity_analysis() |
SCI under alternative weight scenarios |
sci_explain() |
Top loss contributors and pillar breakdown |
Integration with r4subprofile
library(r4subprofile)
library(r4subscore)
prof <- submission_profile("FDA", "NDA")
cfg <- profile_sci_config(prof)
pillar_scores <- compute_pillar_scores(ev, config = cfg)
sci <- compute_sci(pillar_scores, config = cfg)