CFE 2019: Start Registration
View Submission - CMStatistics
B1144
Title: Continuous time hidden Markov models for astronomical gamma-ray light curves Authors:  Andrea Sottosanti - University of Padua (Italy) [presenting]
Mauro Bernardi - University of Padova (Italy)
Alessandra Rosalba Brazzale - University of Padova (Italy)
Luis Campos - Harvard University (United States)
Aneta Siemiginowska - Harvard University (United States)
David van Dyk - Imperial College London (United Kingdom)
Abstract: The detection and characterisation of celestial objects is an inherently inter-disciplinary field which embraces both statistical and astronomical methods. Pioneering technology has driven remarkable acceleration in the rate of detection of celestial objects, and global space astrometry missions will produce accurate maps by surveying stars at an ever-increasing rate. This rapid progression through technology has determined a paradigm shift in observational astronomy, where large digital sky surveys are becoming the dominant source of data. At the same time, astrostatistics has rapidly evolved during the past two decades into a stand-alone discipline with its own professional associations. Indeed, the simple collection of rich, massive data sets on the terabyte scale is not the end of the process but just its beginning. Beyond the technological improvements, a key step in astronomical breakthrough research is the meaningful statistical analysis of the collected information. We introduce a novel approach for the analysis and characterisation of the light curves emitted by time-varying high-energy astronomical phenomena based on continuous time hidden Markov models. The proposed method analyses the variation of signal from a source in time and successfully identifies different latent states that correspond to distinct physical mechanisms. We provide also a bootstrap procedure to evaluate the nature of extreme values in the light curve.