Global and local tests to assess stationarity of Markov transition models

ABSTRACT

We present global and local likelihood-based tests to evaluate stationarity in transition models. Three motivational studies are considered. A simulation study was carried out to assess the performance of the proposed tests. The results showed that they present good performance with the control of the type-I error, especially for ordinal responses, and control of the type-II error, especially for the nominal case. Asymptotically they are close to the classical test performance. They can be executed in a single framework without the need to estimate the transition probabilities, incorporating both categorical and continuous covariates, and used to identify sources of non-stationarity.

KEYWORDS:

• Generalized linear models
• Longitudinal categorical data
• Simulation study
• transition probabilities

MATHEMATICS SUBJECT CLASSIFICATION:

• 62F03
• 62J12

Additional information

Funding

This work was supported by São Paulo Research Foundation (FAPESP), Brazil, grant number 2015/02628-2.