Oura Ring Gen 4 sensor data, not clinical measurementsN=1 case study, not validated for clinical decisionsHEV diagnosed Mar 18; Day 205 post-ruxolitinibMore
Consumer wearable data can support exploratory review only. The HEV diagnosis, temporally confounded with treatment start, remains a material confounder.

Placebo Calibration Tests

20 falsification tests across the pre-treatment period (2026-01-08 to 2026-03-15)
Generated 2026-10-07 06:22 · Model calibration check · Data Jan 08, 2026 to Oct 06, 2026 · Day 205 post-treatment
HRV (RMSSD) FPR
Watch
100%(20/20)
Liberal (FPR=100%, binom p=0.000)
Mann-Whitney false positive rate at alpha=0.05
Lowest HR FPR
Watch
100%(20/20)
Liberal (FPR=100%, binom p=0.000)
Mann-Whitney false positive rate at alpha=0.05
Average HR FPR
Watch
100%(20/20)
Liberal (FPR=100%, binom p=0.000)
Mann-Whitney false positive rate at alpha=0.05
Sleep Efficiency FPR
Watch
60%(12/20)
Liberal (FPR=60%, binom p=0.000)
Mann-Whitney false positive rate at alpha=0.05
SUMMARY

Calibration Summary

Mean false positive rate across metrics: 90.0%. This is substantially above the nominal 5% level. The model is overconfident. Real treatment-effect p-values should be interpreted very cautiously, and stricter significance thresholds (e.g., p < 0.01) may be appropriate.
DETAIL_TABLE

Placebo Test Results (All Dates)

Placebo DateDay #HRV (RMSSD) pHRV (RMSSD) CI pLowest HR pLowest HR CI pAverage HR pAverage HR CI pSleep Efficiency pSleep Efficiency CI p
2026-01-2315p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.066N/A
2026-01-2416p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.028N/A
2026-01-2618p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.023N/A
2026-01-2719p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.018N/A
2026-02-0124p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.039N/A
2026-02-0528p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.035N/A
2026-02-0629p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.027N/A
2026-02-0730p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.049N/A
2026-02-0831p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.084N/A
2026-02-0932p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.195N/A
2026-02-1235p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.101N/A
2026-02-1437p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.041N/A
2026-02-1538p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.017N/A
2026-02-1639p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.012N/A
2026-02-1841p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.044N/A
2026-02-1942p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.102N/A
2026-02-2144p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.026N/A
2026-02-2245p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.066N/A
2026-02-2649p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.076N/A
2026-02-2851p<0.001N/Ap<0.001N/Ap<0.001N/Ap=0.232N/A
FPR_SUMMARY

False Positive Rate Summary

MetricSignificantTotalFPRAssessmentCI SigCI TotalCI FPR
HRV (RMSSD)2020100%Liberal (FPR=100%, binom p=0.000)00N/A
Lowest HR2020100%Liberal (FPR=100%, binom p=0.000)00N/A
Average HR2020100%Liberal (FPR=100%, binom p=0.000)00N/A
Sleep Efficiency122060%Liberal (FPR=60%, binom p=0.000)00N/A
PVALUE_DIST

P-Value Distribution Under Null

Under correct calibration, p-values should be approximately uniformly distributed (flat histogram). A spike near 0 suggests the model is liberal.
METHODOLOGY

Methodology

Purpose: Placebo (falsification) tests check whether the statistical methods used for treatment-effect estimation produce false positives at the expected nominal rate. If they do, the p-values from the real analysis are trustworthy.

Method: 20 random dates were drawn (seed=42) from the pre-treatment period (2026-01-08 to 2026-03-15), each at least 14 days from the window edges. At each placebo date, the pre-treatment data was split and a two-sided Mann-Whitney U test was performed for each metric. CausalImpact (Bayesian structural time series) was also run where available.

Expected result: ~5% of placebo tests should be significant (1 out of 20). If the observed FPR is much higher, the real treatment-effect p-values may be overconfident.

Interpretation:

  • Well-calibrated: FPR within ~2 percentage points of 5%
  • Conservative: FPR near 0% (tests are too strict, may miss real effects)
  • Liberal: FPR significantly above 5% (tests find "effects" where none exist)
Data window: January 08 to October 06, 2026 · Post-treatment: day 205 on ruxolitinib · Generated: 2026-10-07 06:22
Oura Ring Gen 4 consumer wearable data; N=1 exploratory case study; not a medical device. MIT License · © 2026 The Educational Equality Institute · Oura® is a registered trademark of Oura Health Oy.
Source code on GitHub · Built with Claude Code · TEEI Health Equity · Companion project: Genomics portal · How this was built · Every number, checked