Chasing a consistent picture for dark matter direct detection searches

In this paper we assess the present status of dark matter direct searches by means of Bayesian statistics. We consider three particle physics models for spin-independent dark matter interaction with nuclei: elastic, inelastic and isospin violating scattering. We briefly present the state of the art for the three models, marginalizing over experimental systematics and astrophysical uncertainties. Whatever the scenario is, XENON100 appears to challenge the detection region of DAMA, CoGeNT and CRESST. The first aim of this study is to rigorously quantify the significance of the inconsistency between XENON100 data and the combined set of detection (DAMA, CoGeNT and CRESST together), performing two statistical tests based on the Bayesian evidence. We show that XENON100 and the combined set are inconsistent at least at the 2σ level in all scenarios but inelastic scattering, for which the disagreement drops to the 1σ level. Second we consider only the combined set and hunt the best particle physics model that accounts for the events, using Bayesian model comparison. The outcome between elastic and isospin violating scattering is inconclusive, with the odds 2∶1, while inelastic scattering is disfavored with the odds of 1∶32 because of CoGeNT data. Our results are robust under reasonable prior assumptions. We conclude that the simple elastic scattering remains the best model to explain the detection regions, since the data do not support extra free parameters. The outcome of consistency tests implies that either a better understanding of astrophysical and experimental uncertainties is needed or the strength of belief in certain data sets should be revised or the dark matter theoretical model is at odds with the data.