Correlated errors reduce evidence for different circular GJ 273 signals at 413.9 and 541.5 days in HARPS and CARMENES

We test whether two instruments support the same circular velocity curves for GJ 273. We use 250 HARPS nights from 2003 to 2016 and 196 CARMENES DR1 nights from 2016 to 2020. The CARMENES table comes from a verified SCI publication. We test reported outer periods of 413.9 and 541.5 days. We fit separate instrument offsets and slopes, the HARPS upgrade offset, and two inner signals. The joint statistic for different outer coefficients is 94.59 with white errors and 15.38 with independent exponential error covariance in each instrument. In 499 simulations of the fitted common-curve model, 17 statistics meet or exceed the observed value. The adjusted simulation tail is 0.036; the binomial 95% interval is 0.020 to 0.054. The chi-square reference tail is 0.004 and gives stronger evidence than this calibration. When both outer periods can vary within the source intervals, the correlated-error statistic falls to 9.93. That period-range fit has no simulation calibration. An exploratory 2000-day control raises the fixed-period statistic to 20.94. The inner signals give fixed-period statistics of 1.88 and 2.11 for two extra coefficients each. These data show that the outer-curve comparison depends strongly on the error and period model. They do not give a robust rejection of all outer-planet models. We supply exact public inputs, code, figures, fit tables, and simulation results.

GJ 273 · Luyten star · radial velocity · CARMENES DR1 · HARPS · circular signals · stellar activity · parametric bootstrap