This article proposes a new generalization of the Multivariate Markov Chains (MMC) model. The future values of a Markov chain commonly depend on only the past values of the chain in an autoregressive fashion. The generalization proposed in this work also considers exogenous variables that can be deterministic or stochastic. Furthermore, the effects of the MMC's past values and the effects of pre--determined or exogenous covariates are considered in our model by considering a non--homogeneous Markov chain. The Monte Carlo simulation study findings showed that our model consistently detected a non--homogeneous Markov chain. Besides, an empirical illustration demonstrated the relevance of this new model by estimating probability transition matrices over the space state of the exogenous variable. An additional and practical contribution of this work is the development of a novel R package with this generalization.
|Publisher||Cornell University (ArXiv)|
|Number of pages||15|
|Publication status||Published - 1 Feb 2022|
|Name||arXiv.org. Statistics. Methodology|