⚡ ADHD Negative Behavior — Bayesian Causal Network (BCN) Dashboard

Judea Pearl SCM | do-Calculus | Backdoor Adjustment | ρ̄=0.30 Cross-Correlation Removal | 20 Interventions | 6 Pathways
Level I–IV Evidence
🧪 Intervention Controls
0 selected
Pearl-Adj. HR
Select interventions
Rel. Risk Reduction
Abs. risk: —
NNT
ARR: —
Stack PNS
n=0 interventions
Min. E-Value
Unmeasured confounding
⚙️ Global Cross-Correlation ρ̄ (Backdoor Adjustment Sensitivity)
ρ̄ =
Oxford CEBM: Level I→A · II→B · III→C. Lower grades excluded from the combined HR.
0.30 (Low ρ̄ = less redundancy; High ρ̄ = more mechanistic overlap)
📊 Overview
🔀 What-If
🔁 If-Not-For
🎯 Sensitivity
💊 Dose-Response
🕸️ Causal DAG
🧬 Scenarios
📚 Evidence
Pearl-Adjusted HR by Intervention (lower = more protective) Sorted by raw HR
Selected Stack — Marginal Contribution to HR Reduction
Pathway-Level Combined HR
What-If Analysis: For each intervention, shows the marginal change in combined HR if that intervention were added to (or removed from) the current selection. Sorted by impact magnitude.
Waterfall: Marginal HR Contribution (Selected Protocol)
InterventionCurrent HR (all)HR if Added/RemovedΔNBO Prob.Priority
If-Not-For (Population Attributable Risk): What would the NBO probability be if a given intervention had NEVER existed (counterfactual Pearl Layer 3)?
InterventionPAR (abs.)PAR%NBO without thisInterpretation
Sensitivity Analysis: How does the combined HR change as cross-correlation ρ̄ and baseline risk assumptions vary? Highlighted row = current ρ̄ setting.
ρ̄ Sensitivity — Combined HR for Selected Protocol
ρ̄Combined HRAbs. RiskRel. ReductionNNTWeight w
E-Value Sensitivity — Unmeasured Confounding Threshold
InterventionHRE-ValueRobustness
Dose-response profiles for key interventions. Sliders in the left panel adjust the effective HR based on dose tier (Low / Optimal / High).
InterventionDose SettingEffective HRDose DescriptionSaturation Notes
🕸️ Directed Acyclic Graph — Pearl SCM Causal Structure
DAG Legend: Blue arrows = direct causal paths (do-calculus identified). Dashed lines = cross-pathway correlations (ρ). Red node = outcome (NBO). Pathway nodes are mediators. Intervention nodes connect to their respective pathway mediator.
Pre-computed optimal clinical stacks. Click a scenario to load it into the selection panel.
Evidence Summary: Raw effect sizes from systematic reviews and meta-analyses (2018–2025). SMD values pre-dating ρ̄ adjustment.
IDInterventionSMDHR (raw)E-ValuePNSLevelKey Citations