Meditation & Paced Breathing
Motivation
Slow, paced breathing near 0.1 Hz is known to reshape heart-rate variability. Which features move, and by how much relative to a healthy baseline? Using the normative reference, this use case scores a real meditation cohort and reads off its HRV signature as quantile z-equivalents.
Features are scored against the pooled healthy prior (nsrdb + nsr2db, 61 715 windows, 360-beat windows with a 120-beat stride) using the quantile z-equivalent $z = \Phi^{-1}(F_{\text{prior}}(x))$.
The meditation cohort
Dataset. PhysioNet Heart Rate Oscillations during Meditation (Peng et al., 1999): 58 records (meditators and comparison groups, record labels C*, Y*, M*, N*, I*), cut into 2 686 windows (test/testdata/meditation/windowed_w360_s120_features.csv).
Method. For each headline feature, take the median across all 2 686 windows and map it through the healthy prior. Every value below is reproducible from the two shipped CSVs; no display or WFDB tools are needed.
Result. The cohort sits above the healthy population on the classic vagal-tone panel:
| Feature | z-equivalent | Reading |
|---|---|---|
mean (Mean IBI) | +0.99 | longer inter-beat intervals (slower HR) |
sdnn | +0.87 | overall variability elevated |
rmssd | +0.84 | short-term vagal tone elevated |
pnn50 | +0.71 | more large beat-to-beat changes |
The longer mean IBI together with elevated pNN50 is the expected fingerprint of a vagally dominant state, consistent with the slowed, deepened breathing of meditation and its respiratory-sinus-arrhythmia coupling (Akselrod et al., 1981), (Task Force, European Society of Cardiology / North American Society of Pacing & Electrophysiology, 1996).
Takeaway
The normative machinery turns "meditation changes HRV" into a quantified, feature-resolved statement: a broad vagal shift, with mean IBI, pNN50, RMSSD, and SDNN all elevated together against the healthy reference. Context, meaning which features move, is the diagnosis.
Reproduce / where the data lives
- Data:
test/testdata/meditation/windowed_w360_s120_features.csv(2 686 windows) scored againstdocs/normative_priors.csv; aggregation is the all-windows median, quantile z-equivalent. - Scoring: identical to the normative use case:
normative_prior(name)followed by $\Phi^{-1}(\mathrm{cdf}(\cdot))$.