3 The doubtful solutions

The observations reinvestigated during this work are limited to the Hipparcos entries for which a problem has been detected in the routine processing. This does not mean that all such solutions are spurious, in fact most of them are correct, but the proportion of unreliable solutions is larger in this group than in the rest of the Catalogue.

The final statistics published in the Catalogue documentation (ESA 1997, Vol. 1) give an order of magnitude of the number of entries to be reinvestigated on a case by case basis. In addition the different categories help understand what kind of problem may be expected in the reprocessing and provide hints to orient the search for new information.

1.

There were 263 entries with no astrometric solution published. Only 10 were not observed because of large errors in the Input Catalogue positions, while 253 solutions were finally rejected as inadequate.

2.

The solutions with a time dependent proper motion (solutions flagged G in the Catalogue) refer probably to astrometric binaries with periods above about 10 years. There are 2622 such solutions.

3.

The stochastic solutions (flagged X in the Catalogue) that is to say published solutions for which it was not possible to find an acceptable single or multiple star solution in reasonable agreement with the random error of the abscissas. While a significant fraction might be short period astrometric binaries, many others could be true double stars with inadequate relative astrometry.

In addition to these broad categories, there are two other indications that can be used to pinpoint questionable solutions:

1.

Under field H29 one finds the percentage of data that has been rejected in order to converge to an acceptable fit. In general this number is below 10% and a larger value is an indication that the solution should be taken with care as a significant number of outliers has been discarded.

2.

The field H30 attempts to quantify the quality of the final fit, when the outliers have been removed. Values larger than 3 or 4 indicate a bad fit to the data.