← Back to the energy summary · Front page
Client name: Nick
A multi-variate analysis puts several predictors into one model at the same time. Each effect is then the change in readiness for that behaviour with the others held steady. Comparing a behaviour's effect alone with its effect alongside the others shows how much of it was really something else.
This covers Oura readiness only. Nick's own 1–3 rating from the Reflections app had too little signal to be worth modelling this way; see the summary.
next_readiness is shifted back one day by qsci/loaders/oura.py, so the
model pairs rows with no further shift.Points of readiness, with 95% interval and p-value. Ordinary least squares, one row per day, the same 243 days in every model.
| Day-before predictor | Alone | Behaviours together | + bedtime |
|---|---|---|---|
| Alcohol | -4.95 [-7.5, -2.4] · p=0.0001 | -3.94 [-6.5, -1.4] · p=0.0026 | -3.21 [-5.7, -0.7] · p=0.0129 |
| Left comfort zone | -3.50 [-5.4, -1.6] · p=0.0003 | -2.85 [-4.7, -1.0] · p=0.0028 | -2.53 [-4.4, -0.7] · p=0.0067 |
| Caffeine, per cup | -1.99 [-3.6, -0.4] · p=0.0158 | -1.40 [-3.0, +0.2] · p=0.0813 | -0.28 [-1.9, +1.4] · p=0.7363 |
| Bedtime, per hour later | -2.70 [-3.8, -1.6] · p=0.0000 | — | -2.13 [-3.3, -1.0] · p=0.0002 |
| Share explained (R²) | — | 0.104 | 0.152 |
Correlations between the predictors on these days. Caffeine and bedtime moving together is why caffeine's effect shrinks once bedtime is in the model.
| Alcohol | Left comfort zone | Caffeine, per cup | Bedtime, per hour later | |
|---|---|---|---|---|
| Alcohol | +1.00 | +0.17 | +0.15 | +0.21 |
| Left comfort zone | +0.17 | +1.00 | +0.09 | +0.14 |
| Caffeine, per cup | +0.15 | +0.09 | +1.00 | +0.39 |
| Bedtime, per hour later | +0.21 | +0.14 | +0.39 | +1.00 |