Responder Floor Atlas

Continuous-vs-responder framing reproducibility on Pairwise70 — generated from commit 83cf940 on 2026-04-25.

Q1 — Framing flip rate

MetricValue95% CI (clustered bootstrap)
Reviews pooled4
Framing flips (α=0.05)1 (25.0%)[0.0%, 75.0%]
Magnitude flips (|Δ logRR|>0.1)4 (100.0%)
0 4 1 flip
1 of 4 reviews flipped

Q2 — Reconstruction fidelity

Instrumentn trialsMedian |ε|95th percentile% |ε|>0.05
bmi_change2360.30170.727693.6%
body_weight_kg5860.25960.639688.7%
body_weight_pct6990.22440.667683.3%
sf36_mcs1720.27380.650291.9%
sf36_pcs1720.34800.585891.9%
sf36_physical_function1480.31290.547194.6%

Distribution of |ε| by instrument

BMI change 0.000 0.01 0.9 Weight (kg) 0.000 0.01 0.9 Weight (%) 0.000 0.01 0.9 SF-36 MCS 0.000 0.01 0.9 SF-36 PCS 0.000 0.01 0.9 SF-36 PF 0.000 0.01 0.9 |ε| (arm-level reconstruction error)

Bland–Altman: reconstructed vs observed responder proportions

+1.96SD=0.744 −1.96SD=-0.496 mean=0.124 -0.86 -0.43 0.00 0.43 0.86 0.00 0.25 0.50 0.75 1.00 Mean of p̂ and p_obs p̂ − p_obs n = 200 (sampled)

Q3 — Implied-MID atlas

Instrumentn trialsMedian implied MID95% range
bmi_change2362.853-38.507–8.712
body_weight_kg5867.995-89.826–23.916
body_weight_pct6996.681-6.188–22.772
sf36_mcs1723.669-14.813–18.858
sf36_pcs1726.591-8.184–20.018
sf36_physical_function1487.728-8.466–22.684

Bootstrap CIs on implied-MID ratio (empirical / canonical)

0.0 2.8 5.5 8.3 11.0 13.8 Implied MID (absolute) BMI change ratio = 2.85 [2.64–3.81] Weight (kg) ratio = 1.60 [1.34–2.29] Weight (%) ratio = 1.34 [1.28–2.55] SF-36 MCS ratio = 1.22 [1.22–1.22] SF-36 PCS ratio = 3.30 [3.30–3.30] SF-36 PF ratio = 1.55 [1.55–1.55] canonical MID

Q4 — Cross-review trial overlap

Of 120 unique trial-instrument combinations, 7 appeared in ≥2 reviews (max overlap = 2 reviews). 113 were unique to a single review.

InstrumentTrialReviews (n)Review IDsδ̂ rangeδ̂ min–max
bmi_changeO'Neil 20182CD015092_pub2, CD0160178.205-1.026–7.179
body_weight_kgO'Neil 20182CD015092_pub2, CD01601723.661-3.987–19.674
body_weight_kgSTEP 82CD015092_pub2, CD01601730.507-12.286–18.220
body_weight_kgSURMOUNT-52CD015092_pub2, CD01601828.8059.033–37.838
body_weight_pctO'Neil 20182CD015092_pub2, CD01601721.043-2.626–18.417
body_weight_pctSTEP 82CD015092_pub2, CD01601726.152-9.967–16.185
body_weight_pctSURMOUNT-52CD015092_pub2, CD01601828.5118.219–36.730

Q5 — Sensitivity: implied MID under alternative distributional assumptions

InstrumentLog-normalTruncated-NormalBeta
Median ΔpP95MaxMedian ΔpP95MaxMedian ΔpP95Max
bmi_change0.02760.04490.05140.00280.00910.01450.00360.22970.5095
body_weight_kg0.03020.04240.05170.00300.00870.01460.00420.01410.5109
body_weight_pct0.02750.04210.04960.00280.00890.01510.00350.01060.0242
sf36_mcs0.02790.04210.04800.00300.00890.01370.23460.36850.3722
sf36_pcs0.03740.04550.05120.00340.00980.01680.32740.39870.4061
sf36_physical_function0.03520.04180.05140.00320.00880.01540.25420.35010.3687

Spec: docs/superpowers/specs/2026-04-22-responder-floor-atlas-design.md