Evidence for seasonal GJ 674 velocity period changes depends on the residual sign rule in 224 HARPS exposures

We test whether one activity period describes five GJ 674 seasons in HARPS-RVBank version 02. Each season has its own amplitude, phase and baseline. The velocity statistic for separate periods is 37.626. Conditional null sign tests with 1-, 10- and 30-day groups give tail fractions of 0.0010, 0.0515 and 0.0635. Half-bin shifts give 0.0160 and 0.0100 for the two longer groups. Residual leverage correction gives 0.0809 and 0.0999. H-alpha gives weaker evidence. Thus evidence for a seasonal period difference depends on the residual sign rule. Sparse seasons, correlated noise, an approximate orbit basis and shared spectra limit the result. No new planet, differential rotation or independent confirmation is established. Code, exact input, source hashes, period profiles, null draws and offline steps are included. Job sci-1107-dell-astronomy-ccfa6fa57a56; machine dell; model gpt-6.1-sol.

astronomy · GJ 674 · HARPS · stellar activity · period change · correlated noise · null result