Bernoulli vector autoregressive model

Abstract In this paper, we propose a vector autoregressive (VAR) model of order one for multivariate binary time series. Multivariate binary time series data are used in many fields such as biology and environmental sciences. However, modeling the dynamics in multiple binary time series is not an easy task. Most existing methods model the joint transition probabilities from marginals pairwisely for which the resulting cross-dependency may not be flexible enough. Our proposed model, Bernoulli VAR (BerVAR) model, is constructed using latent multivariate Bernoulli random vectors. The BerVAR model represents the instantaneous dependency between components via latent processes, and the autoregressive structure represents a switch between the hidden vectors depending on the past. We derive the mean and matrix-valued autocovariance functions for the BerVAR model analytically and propose a quasi-likelihood approach to estimate the model parameters. We prove that our estimator is consistent under mild conditions. We perform a simulation study to show the finite sample properties of the proposed estimators and to compare the prediction power with existing methods for binary time series. Finally, we fit our model to time series of drought events from different regions in Mexico to study the temporal dependence, in a given region and across different regions. By using the BerVAR model, we found that the cross-region dependence of drought events is stronger if a rain event preceded it.

[1]  J. Teugels.  Some representations of the multivariate Bernoulli and binomial distributions , 1990 .

[2]  A. Raftery,et al.  The Mixture Transition Distribution Model for High-Order Markov Chains and Non-Gaussian Time Series , 2002 .

[3]  Babak Shahbaba,et al.  Modeling Binary Time Series Using Gaussian Processes with Application to Predicting Sleep States , 2018, J. Classif..

[4]  Brendan McCabe,et al.  Forecasting discrete valued low count time series , 2004 .

[5]  Michael L. Stein,et al.  A stochastic space-time model for intermittent precipitation occurrences , 2016, 1602.02902.

[6]  Adelchi Azzalini,et al.  Logistic regression and other discrete data models for serially correlated observations , 1994 .

[7]  Patrizia Semeraro,et al.  Characterization of multivariate Bernoulli distributions with given margins , 2017 .

[8]  G. Wahba,et al.  Multivariate Bernoulli distribution , 2012, 1206.1874.

[9]  A. Raftery.  A model for high-order Markov chains , 1985 .

[10]  Rob J Hyndman,et al.  Nonparametric additive regression models for binary time series , 1999 .

[11]  João Nicolau.  A New Model for Multivariate Markov Chains , 2014 .

[12]  João Nicolau,et al.  A simple nonparametric method to estimate the expected time to cross a threshold , 2017 .

[13]  D. Cox,et al.  A note on pseudolikelihood constructed from marginal densities , 2004 .

[14]  Anne-Catherine Favre,et al.  The new family of Fisher copulas to model upper tail dependence and radial asymmetry: Properties and application to high‐dimensional rainfall data , 2018 .

[15]  C. Varin,et al.  A mixed autoregressive probit model for ordinal longitudinal data. , 2010, Biostatistics.

[16]  J. C. van Houwelingen,et al.  Logistic Regression for Correlated Binary Data , 1994 .

[17]  Claudia Czado,et al.  Predictive Model Assessment for Count Data , 2009, Biometrics.

[18]  Ed. McKenzie,et al.  SOME SIMPLE MODELS FOR DISCRETE VARIATE TIME SERIES , 1985 .