Title: An extension of multivariate models for a monotone disease under misspecification
Authors: Ines Varas - Pontificia Universidad Catolica de Chile (Chile) [presenting]
Abstract: Caries experience (CE) is an essential topic in dental and medical sciences. This process is prone to misclassification because of several reasons. The CE process is a monotone disease because when a tooth with caries has been found, it can reverse its state. Hidden Markov models (HHMs) have been applied in several research areas to connect latent and observed states of a process along time. These models allow the definition of different types of misclassification structures. Most of the research already developed in this framework considers a dichotomic definition of both the latent and observed processes. Nevertheless, the definition of two stages processes could be restrictive for modelling the CE process in the sense that the model does not describe the intermediate stages of it. We propose a hidden Markov model for a misclassified longitudinal monotone process defined in more than two categories to evaluate the prevalence and incidence of the process. Identifiability restrictions to obtain consistent and interpretable parameters are evaluated.