Parameter-driven models for time series of count data

TL;DR

This paper evaluates parameter-driven models for count time series, comparing estimators like GLM, finite mixture, and hidden Markov models through simulations and real data applications, highlighting their accuracy and robustness.

Contribution

It provides a comprehensive simulation study and practical insights into the performance of various estimators for count time series models.

Findings

GLM estimators are generally efficient and robust.

HMM estimators are suitable only in extreme cases.

Standard errors for estimators are derived and validated.

Abstract

This paper considers a general class of parameter-driven models for time series of counts. A comprehensive simulation study is conducted to evaluate the accuracy and efficiency of three estimators: the maximum likelihood estimators of the generalized linear model, 2-state finite mixture model, and 2-state hidden Markov model. Standard errors for these estimators are derived. Our results show that except in extreme cases, the maximum likelihood estimator of the generalized linear model is an efficient, consistent and robust estimator with a well-behaved estimated standard error. The maximum likelihood estimator of the 2-state hidden Markov model is appropriate only when the true model is extreme relative to the generalized linear model. Our results are applied to problems concerning polio incidence and daily numbers of epileptic seizures.