B0827
Title: Nonparametric modelling of periodic variation in hidden Markov models
Authors: Carlina Feldmann - Bielefeld University (Germany) [presenting]
Sina Mews - Bielefeld University (Germany)
Roland Langrock - Bielefeld University (Germany)
Abstract: Within the class of hidden Markov models (HMMs), a popular tool for modelling time series driven by underlying states, periodic variation in the state-switching dynamics is routinely modelled using trigonometric functions. This parametric modelling can be too inflexible to capture complex periodic patterns, e.g. featuring multiple activity peaks per day. The alternative approach uses cyclic penalised splines to model periodic variation within HMMs. The challenge of estimating the corresponding complex models is substantially reduced by the expectation-maximisation algorithm, which allows the use of the existing machinery (and software) for nonparametric regression. This approach's practicality and potential usefulness are demonstrated in a real-data application modelling the activity of fruit flies.
