At fixed geometry, 0.25 m/s correlated noise with a 15-minute time scale cuts the HD 3167 b chi-square gain by 48%

We test 254 public velocities from five transits of HD 3167 b. Winn et al. (2026) found a common transit signal with independent measurement errors. We fix their central transit geometry and fit two small-rotation velocity templates. Each night has a free offset and slope. The chi-square gain is 21.286 with the published errors. An added exponential covariance with amplitude 0.25 m/s and time scale 15 minutes reduces the gain to 11.106. The full nonlinear velocity model gives similar gains of 21.615 and 11.290. The 161 velocities outside transit permit this noise level. At a fixed 15-minute time scale, their best amplitude is 0.256 m/s. They do not establish a unique noise process. A search that fits the covariance under both hypotheses gives a likelihood-ratio statistic of 12.527. Under the fitted null covariance, 225 of 20,000 simulations exceed the corresponding finite-grid statistic. The tail estimate is 1.13%, with a binomial 95% interval of 0.983% to 1.281%. This interval includes simulation sampling error only. The per-night coefficients remain statistically consistent with a common signal. The result measures sensitivity to a stated noise family. It does not give a new planet detection or a new orbital angle. We preserve the source table, data provenance, code, all tested settings, simulation draws, figures, and reproduction commands.

HD 3167 b · Rossiter-McLaughlin effect · radial velocity · correlated noise · exoplanets · reproducible research