|
Abstract:
|
The Large Hadron Collider (LHC) at CERN is set to produce data that may shed light on the existence, or otherwise, of the Higgs-Boson particles. One type of data to be collected may be modeled using a system of three Poisson models corresponding to experimental observation, in addition to instrumental and background calibrations. Several Bayesian approaches are explored to provide marginal posterior percentiles for the parameter of interest. With multiple decay channels, dimensionality of the nuisance parameter grows and specification of the prior structure becomes increasingly challenging. Problems with vague priors in high-dimensions are illustrated in the multichannel case. A systematic approach is taken using single-level and hierarchical models, with full Bayes and Empirical Bayes methods compared. The goal is to provide upper limits with excellent Frequentist coverage properties.
|
