Statistical Methods for classification of the XMM-Newton X-Ray Sources

The XMM-Newton Survey Science Center (SSC) XID Project has, as the goal of its core programme, the statistical identification of the entire database of serendipitously observed XMM-Newton X-ray sources. To this end, the programme has sought to identify, through optical/IR follow-up photometric and spectroscopic observations, the optical counterparts of a significant subset of the X-ray sources. The aim is to calibrate the properties of this identified subset and propagate it to the wider XMM-Newton source catalogues, yielding a statistical identification for the hundreds of thousands of serendipitously detected objects for which individual follow-up identification observations would be impractical. The repository for the resulting data, the XID Database, permits extensive exploitation and analysis of the information. Here We show that up to statistical uncertainties, each class of X-ray sources has specific photometric - optical/IR and X-ray - characteristics, and therefore it is possible to use them for statistical classification of sources. We investigate a number of statistical methods and examine their relative and absolute performance by comparing the behaviour of physical quantities for statistically classified objects with what is obtained from spectroscopy. We conclude that among methods we have studied, multi-dimensional probability distribution is the best for both classifying source type and redshift. We also discuss the effect of classification method and the input set on the astronomical conclusions about distribution and properties of the X-ray selected sources.