Every number on your report is traceable to the sessions that produced it.
Axon is a measurement instrument, not an oracle. This page is the full path from a logged dose to an effect size — including every gate where the engine refuses to print a number the data hasn't earned.
From log to effect size, with nothing generative in between.
Inputs. Doses, 0–100 visual-analogue ratings at pharmacokinetically timed checkpoints, optional external task scores, and context — sleep, stress, time of day. Everything you log; nothing you don't.
Engine. Dose curves become exposure series. Within a session, an AR(1) dynamic linear model relates exposure to your ratings. Across sessions, effect sizes are Hedges-corrected Cohen's d against your own unmedicated baseline. Deterministic: same logs in, same numbers out, every time.
Outputs. Effect sizes with confidence flags, a structured readout, and a full export of everything you've logged. The optional AI layer writes prose around the numbers — it can never change them.
Where the engine says “not yet.”
The hardest part of n-of-1 statistics isn't producing a number — it's declining to. Four gates stand between your logs and your report.
Observations farther than 3 median-absolute-deviations from your own median are excluded as likely logging errors — and every exclusion is listed on the insights page, not silently dropped.
When the data is too sparse for the posterior to escape its skeptical prior, the number is suppressed entirely. Axon would rather show you nothing than show you the prior dressed up as a finding.
Compounds that act over weeks don't produce within-session dynamics, so the within-session model returns no estimate — by design. No estimate means no number, not a fabricated one.
Early signals from fewer than five sessions are shown, but carry an explicit experimental flag until your sample stabilizes. The flag is removed by data, never by time.
Yours, in a format you can leave with.
Every session, rating, and score is exportable in one download. Nothing is shared unless you explicitly choose to share it. Axon's value is the rigor of the analysis, not a lock on your logs.
An instrument, not an advocate.
Axon measures what you choose to investigate. It does not recommend compounds, suggest doses, or make health claims — and an n-of-1 trial describes you, not anyone else. It is not a diagnostic tool and not medical advice. Read the full disclaimer.