Reliability in Multilevel Continuous Time Modeling

Google map

Abstract:

Continuous time (CT) models have become increasingly prevalent in recent years. In contrast to traditional discrete time (DT) models, CT models can readily handle observations measured at unequal intervals and offer the flexibility to explain dynamic and cross-process coupling effects irrespective of the specific lengths of time intervals. Despite their growing popularity, discussion of issues of reliability within the CT framework is sparse. Reliability, the consistency of measurement over repeated tests, is crucial for ensuring similar results across replications. In multilevel longitudinal data, questions of reliability encompass both within-person (consistency over time for the same individual) and between-person (consistency across different individuals) considerations within the DT modeling framework. This study proposes and evaluates extensions of the reliability indices developed for the DT framework to the CT framework. And we explore the properties of these reliability indices for multilevel irregularly spaced longitudinal data in a Monte Carlo simulation study.