Hidden Markov Models

Hidden Markov model (HMM; also known as latent Markov model) is an extension of the Markov model to applications where the states in the assumed Markov process are hidden or latent. This family of models is generally used to model a sequence of data—such as panel data—when the state produced by the stochastic process cannot be observed directly.

Panel data are used in many disciplines such as education, sociology, marketing, economics, medicine, and psychology. Traditionally, there have been three approaches to investigate change in a variable measured at multiple occasions. These three are (a) marginal or population-averaged models, (b) random-effects or growth models, and (c) transitional models. All three differ not only in the assumptions but also in the research questions they shed light on. These ...