You change something — magnesium before bed, no caffeine after noon, an earlier lights-out — and a few days later you feel… maybe better? A self-experiment turns that hunch into a real answer. Done right, an n-of-1 test compares a stretch of days with the change against your own baseline before it, on the metrics that actually respond, and tells you whether the effect is real or just noise.
n-of-1 means one subject: you. Instead of averaging an effect across hundreds of strangers, you use yourself as your own control — comparing a period with the change against a period without it. For a personal decision ("does magnesium help my sleep?") that's the most relevant evidence there is, because it's measured on the one body you care about. The catch is that a single person's data is noisy, so the method matters.
Change one variable at a time. If you start magnesium, shift your bedtime and cut caffeine in the same week, you'll never know which one did anything. Then pick an outcome plausibly linked to it: magnesium and an earlier bedtime plausibly show up in deep sleep and HRV; a late meal or a nightcap show up in overnight resting heart rate; caffeine shows up in sleep latency. Testing a change against a metric it can't affect just produces a shrug.
HRV, resting heart rate and sleep stages swing from night to night for reasons that have nothing to do with your experiment — a hard workout, a glass of wine, a stressful day. A two-day trial measures that noise, not your change. Give each phase roughly two to four weeks so a genuine shift has time to separate from the daily scatter, and try to keep the obvious confounders — illness, travel, a heatwave — out of the comparison.
Three in particular. Regression to the mean: you usually start an experiment when you feel bad, so you'd have drifted back toward normal anyway — and you'll credit the pill. Placebo: expecting a change quietly shifts how you feel, which is exactly why an objective number like HRV is worth more than "I think I slept better." And confounders: a work deadline or a heatwave can move your metrics further than your intervention ever will. You can't eliminate these, but a long enough window and an honest baseline blunt them.
The only fair comparison is against your normal. If your HRV usually swings between 45 and 65 ms, an average of 58 during your magnesium month is inside your own noise — not a result, however much you want it to be. A real effect is one that clears your personal variation, not one that beats a population average you found in an article. That's the difference between a story you tell yourself and something you can actually act on.
Vitra has a self-experiment tool built for exactly this. You name the variable you're testing and mark when you start; Vitra reads your Oura ring on your own machine and compares the after-period against your own before-baseline across HRV, sleep and resting heart rate, then tells you whether the change cleared your normal variation or sat inside it. Everything is computed on-device against your own history — no cloud, no AI model, and no pretending noise is a breakthrough.
Local AI on your Mac or PC. One-time purchase, 7-day trial, no Vitra subscription.
Download Vitra →