23 policy scenarios ranked - bars = mean, whiskers = 95% CI, *** = significant vs Baseline (Holm), dashed = Baseline
Is the model trustworthy? Every output lands inside its real-world plausible range - green band = literature range, dot = model (all 6 PASS)
Cost-effectiveness - cheaper (left) & more effective (up) is better; shaded corner beats Baseline; dashed = Baseline; hover a dot for its scenario
More locals per park → higher retention: each extra local keeps ~4 more people in 100 (shaded = 95% confidence)
Gender equity - female-male dropout gap (pp): dot = mean, line = 95% CI · coral = real gap (CI excludes 0), grey = n.s. · 0 = parity
Buddy programme - retention by pairing window: week-8 (coral) beats week-0 (steel); degree ≈ random; dashed = Baseline
Scenario
Play the programme year (week 0 → 52)
Macro dynamics (300-run mean ± SD) - colour = selected scenario, dashed grey = Baseline, dashed vertical = winter onset (wk 9)
Baseline programme year - true spatial replay (seed 1, the canonical run)
World view: ▲ park · ★ trainer · ● migrant (red) / local (blue), brightness = motivation · grey = dropped · orange = winter-paused · orange link = cross-group tie
All 23 scenarios (pinned values from the verified pipeline)
Provenance - every number traces to a pinned table; the dashboard computes no new statistics
  • Ranking / cost / tie / CEFR / motivation: tables/table_cost_effectiveness.csv + tables/table2_descriptive.csv (pinned v6.4; Baseline retention 45.1%, cross-tie 0.502, CEFR 1.05, cost €3155 - matching the thesis validation table).
  • Significance (***): tables/table3_hypothesis_tests.csv, Holm-corrected, retention vs Baseline.
  • Dose-response: tables/table_dose_response.csv. Gender gap: tables/table_equity_gender.csv. Validation: tables/table1_validation.csv (all 6 PASS).
  • Dynamics & replay: per-run data/<scenario>/CIM_timeseries_*.csv (300-run mean ± SD) and CIM_edges_Baseline_1.csv / the seed-1 spatial export.
  • 95% CI = mean ± 1.96·SE, SE from each table’s reported SD/n. The NetLogo model is unchanged and bit-for-bit reproducible.
  • Synthetic simulation. All figures describe the behaviour of a calibrated agent-based model, not real individuals; no personal data is used or shown.