Reviewed public evidence · N-of-1
Computational evidence
Models test whether governed movement states are distinguishable under specified training and holdout rules. The approved mechanical-state synopsis supports separation within this one-person record, while negative and null checks constrain interpretation. Model accuracy is not diagnostic accuracy, an estimate of pain, or a population-wide validation result. Each model belongs to a specific question, input cohort, feature set, and evaluation design. The existing graph registry retains historical and exploratory models with their original status labels. Use the detailed tables and provenance to see which results were held out, which were descriptive, and where a finding did not generalize.
Summary and explore
Use the reviewed graph registry and the methods below for this topic.
| Question | Input | Sample | Validation | Result | Limitation |
|---|---|---|---|---|---|
| Can governed Mall and PT mechanical states be distinguished? | Fixed triplet mechanical features | One participant; grouped triplet protocol | fixed triplet-grouped cross-validation | 0.98 balanced accuracy | N-of-1 repeated observations; models reuse the same underlying protocol and are method convergence rather than independent replication. R² is held-out fit, not a clinical outcome. Pain is first-person evidence. |
| Did the mechanical direction repeat on a separate five-date holdout? | Held-out governed triplets | 5 | No date overlap with the training period | 5 of five dates | This checks direction within the same N-of-1 protocol, not population generalization. |
| Do unsupervised clusters recover pain-context labels? | Label-blind time windows | 30-second and 60-second analyses | Adjusted Rand agreement after revealing labels | -0.018953013054371238 30-second adjusted Rand index | Label-blind clusters had near-zero adjusted Rand agreement with revealed pain-context labels. PC-01 also found within-context pain prediction unidentifiable. |
Reviewed reference values
Explore headline values and their units
- Mall/Walking ratio of medians
- 14.1728 ×
- Historical transport rides
- 59 rides
- Mall skating median distance
- 2.0026 miles
- Mall vertical RMS
- 0.05199422678336572 g
- Matched dates
- 55 dates
- May 23 late sustained mechanical episode
- 24 minutes
- May 23 subsequent HR change
- 10.0946 bpm
- May 23 subsequent SDNN change
- 54.2489 ms
- Corrected October 2 duration
- 40:50 mm:ss
- October 2 duration percentile
- 64.4068 empirical percentile
- October 2 corrected jerk
- 6.1269 g/s
- October 2 corrected jerk above historical one-minute P75
- 35.7436 observed minutes
- October 2 corrected jerk above historical one-minute P90
- 16.9103 observed minutes
- October 2 corrected jerk percentile
- 93.2203 empirical percentile
- Walking higher RMS dates
- 55 of 55 dates
- Walking median distance
- 0.1413 miles
- Walking vertical RMS
- 0.13820607087274947 g
- ParaTransit WHOOP dates
- 112 dates
- Activity means ≥100 bpm
- 33 sessions
- Median activity mean HR
- 87.0 bpm
- ParaTransit WHOOP sessions
- 161 sessions
Methods and provenance
Methods and provenance
Publication status: REVIEWED_PUBLIC_SITE_PROJECTION. Latest reviewed observation: 2026-10-02. The legacy graph registry retains its separate 2026-09-17 cutoff.
Reviewed resource hash: 9515359e327b17542d7ae2d008a372916245ac610b8aa2cd4f5987a3e4e8a6cb. Topic projection hash: d773b8e41e014000e9a6725adcc20524090214b17543e1243da049533c1b9bd6.
Source runs: 20261005T-VIS-MOB-01-V2, 20261005T-TRANSPORT-BURDEN-01-V3, 20261005T-TRANSPORT-PHYSIO-MECH-01-V5, 20261005T-TRANSPORT-CONTEXT-01-V5, TRANSPORT-HR-CONTEXT-01-V1, 20261002-v3-reference-v1, MOBILITY-200HZ-01.
Measurements describe observed function, mechanics, physiology and exposure at distinct grains. They do not directly measure pain or internal joint force.
Publication manifest