Statistical properties of Hipparcos 2, caveats on its use, and a recalibration of the Intermediate Astrometric Data

We revisit the Hipparcos 2007 re-reduction and find improvements to the catalog by leveraging Gaia EDR3. We show that including a constant residual offset and additional dispersion (two free parameters in total) in the Hipparcos 2007 IAD creates a new catalog with significantly better agreement with Gaia EDR3. The astrometric parameters, after recalibration, have z-scores that follow a unit-Gaussian when measured against Gaia EDR3 values. We have expanded the Python astrometry tool, htof, to recalibrate the IAD on-the-fly. On a second front, we find that a merged set of IAD from the 1997 and 2007 Hipparcos reductions is not possible in an internally consistent manner. This can be understood if Hipparcos 2007 is an improved, but overfit, model to the underlying along-scan data. For this reason, we recommend using the recalibrated Hipparcos 2007 astrometric parameters, or those from the Hipparcos-Gaia Catalog of Accelerations – because the signatures of overfitting are calibrated out. We advise caution in fitting orbits to the IAD from either Hipparcos 2 as-published or the recalibrated version presented here.