EPQ — Lorenz sensitivity

Unprocessed output

Raw simulation data

Eight uncut member runs from LorenzLyapunov.java. Each file holds 50,001 points at dt = 0.001 and is preceded by its configuration header, so the settings travel with the data.

How to verify a result. Download any file, then load it into the simulation page’s Java–CSV comparison. The browser applies the same stated fit rule to the supplied data and shows both exponents side by side.
Ensemble mean λ = 0.833 ± 0.124 — 0.59 standard deviations from Sprott’s published 0.906.
MemberSpin-upReported λRaw CSV
1300.972Download CSV
2350.853Download CSV
3400.910Download CSV
4450.827Download CSV
5500.866Download CSV
6550.681Download CSV
7600.615Download CSV
8650.937Download CSV

The spread across members is the honest uncertainty in this measurement: the same program, the same rule, eight different starting points on the attractor. The Java source and instructions are on the supervisor and assessor page.