A1435
Title: Count time series: Handling structural breaks
Authors: Isabel Pereira - University of Aveiro (Portugal) [presenting]
Magda Monteiro - University of Aveiro (Portugal)
Ines Pinto - University of Aveiro (Portugal)
Abstract: Count time series are commonly used in fields like healthcare, finance, and transportation to track events such as hospital admissions or stock transactions. A well-known approach for modeling stationary count data is based on thinning operators and mimics autoregressive moving-average models. This method can generate stationary integer-valued time series with distributions like negative binomial, geometric, or Poisson. One widely studied model is the INAR(p) model, where p represents the autoregressive order. However, these models typically assume constant parameters over time, which may not always hold in practice. For instance, during an epidemic, the number of daily cases typically begins low, rises, and eventually falls. Identifying breakpoints in count processes is essential for both methodological and practical purposes, such as validating scientific hypotheses, overseeing safety-critical systems, and ensuring the accuracy of model assumptions. Considering the INAR model with structural breaks and an innovation process that accounts for overdispersion, both classical and Bayesian approaches are addressed to estimate the model parameters. The proposed methodologies are applied to a real dataset in the context of health indicators. Various techniques for detecting one or more change points are compared, utilizing both frequentist and Bayesian frameworks.
