Hierarchical hidden Markov models: applications in ecology and economics
Speaker: Timo Adam (CREEM, University of St.Andrews.)
Hidden Markov models constitute a versatile class of statistical models for time series where the observed variables are driven by hidden states. While basic hidden Markov models are restricted to modelling single-scale data, different variables are often observed at different temporal resolutions. Distances travelled by an animal, for instance, are often obtained from GPS tags on an hourly basis, whereas accelerations obtained from accelerometers are available at much higher sampling frequencies, with observations typically made several times per second. This seminar provides a gentle introduction to hierarchical hidden Markov models, which allow to model such multi-scale time series by regarding the observations as stemming from multiple, hierarchically structured state processes, each of which operates at the time scale at which the corresponding variables were observed. The suggested approach is illustrated using case studies from ecology and economics.