Coherent noise at 25.896 days lowers twice the log likelihood gain from 41.05 to 21.80 in a test of 207 HD 11506 velocities

We test the reported 25.896-day radial velocity candidate in HD 11506 with 207 public measurements. We fix the candidate period and conjunction phase. We refit two known orbits, a quadratic trend, three instrument offsets and all free noise amplitudes under both hypotheses. Across 15 preset covariance shapes, twice the fitted log likelihood gain ranges from 21.80 to 44.40. Candidate amplitudes range from 3.52 to 3.80 m/s. Independent noise gives a gain of 41.05. A Gaussian process at the candidate period, with infinite coherence and fixed unit smoothness, gives 21.80. Its null correlated amplitude is 2.54 m/s. With the candidate present, that amplitude reaches zero and the candidate amplitude is 3.71 m/s. All ten planned trend and instrument checks retain a positive candidate amplitude. Three later smoothness checks give gains from 20.69 to 30.25. Amplitude profiles show that repeating noise can absorb some mismatch away from the best candidate amplitude. These are fitted likelihood comparisons, not calibrated false-alarm probabilities or planet probabilities. The same velocities helped to select the candidate in earlier work. The tests cover fixed covariance shapes and do not use activity indicators. The cause of the variation remains unknown. The package preserves code, source URLs and hashes, full fit solutions, numerical checks, figures and run instructions.

HD 11506 · radial velocity · planet candidate · correlated noise · Gaussian process · reproducible research