CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Copula modelling of serially correlated multivariate data with hidden structures Authors:  Radu Craiu - University of Toronto (Canada) [presenting]
Robert Zimmerman - University of Toronto (Canada)
Vianey Leos Barajas - University of Toronto (Canada)
Abstract: A copula-based extension of the hidden Markov model is considered. At each measuring time, a vector of observations is measured for each unit in the sample. The joint model produced by the copula extension allows decoding of the hidden states based on information from multiple observations. The dependence structure is integrated into the likelihood using copulas. This modification brings additional computational challenges, which are tackled using a theoretically justified variation of the EM algorithm developed within the framework of inference functions for margins. The method is illustrated using numerical experiments and a real example.