McMaster University
Multivariate Gaussian Hidden Semi-Markov Models for Longitudinal Data
Abstract
dc:description.abstractHidden semi-Markov models (HSMMs) represent a sequence of latent states, each of which influences the observable outcomes. Since HSMMs incorporate the explicit specification of state duration distributions, they allow for more flexible modelling of the time spent in each state and are especially suitable for situations in which the hidden Markov model’s assumption of geometrically distributed state durations is not appropriate. Despite a growing body of literature on HSMMs, they are a relatively recent approach in longitudinal data analysis, and both theoretical and applied papers in this context are scarce. As further research is conducted in this area, over-parametrization may become a significant challenge. In model-based clustering methods and hidden Markov models (HMMs), parsimony has been introduced via constraints on the eigen-decomposed covariance matrices. Following this approach, a family of HSMMs for longitudinal data with multivariate Gaussian emission distributions, referred to as the EDG-HSMM family, is proposed. The performance of the family is illustrated using several simulated and empirical datasets.
Degree
thesis:*- Department dc:contributor.department
- Mathematics and Statistics
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Adalyn de Grosbois
- Advisor dc:contributor.advisor
-
- McNicholas, Paul D.
Subjects
dc:subject × 5Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-NoDerivatives 4.0 International
- Licence dc:rights.uri
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Identifier URI
- https://doi.org/10.71548/7682