A quadratic Earth-velocity model does not reduce held-year error in 169 GJ 163 HARPS extraction differences

We test whether Earth velocity predicts the difference between two public HARPS velocity reductions of GJ 163. We use 169 accepted exposures before the 2015 instrument change. A constant and time trend form the base model. A linear and squared Earth-velocity term raise squared error by 0.247 percent when each of 11 year bins is held out. A conditional year-resampling gain range is -12.40 to +4.16 percent. Four residual sign tests give upper-tail fractions from 0.141 to 0.363. All nine controls have gain ranges that include zero. Thus these tests do not support the specified stable correction. Conditional signal trials show sensitivity to exact added shapes, but do not give a general exclusion limit. Shared spectra, sparse years, error correlation and restricted model shapes limit the result. No new planet, unique physical cause or independent instrument confirmation is established. Original code, aggregate results, source hashes and retrieval steps are included. External measurements stay outside the signed package.

astronomy · GJ 163 · HARPS · Earth velocity · spectral extraction · cross validation · null result