A1355
Title: A pairwise likelihood framework for multivariate ordinal time series
Authors: Anna Nalpantidi - Athens University of Economics and Business (Greece) [presenting]
Dimitris Karlis - RC Athens University of Economics and Business (Greece)
Abstract: Multiple dependent ordinal time series, such as unemployment levels across different countries, require appropriate multivariate modeling approaches. However, specifying their joint distribution is challenging, and this difficulty is compounded in time series contexts where both serial and cross-dependence must be considered. A copula-based approach offers a solution for modeling the joint distribution of ordinal variables. For bivariate cases, each pair can be specified using an appropriate copula. In higher-dimensional settings, the full likelihood can be approximated using composite likelihood methods, specifically pairwise likelihood, where each bivariate model is specified by a copula. Rather than maximizing the pairwise likelihood directly, individual bivariate models can be maximized independently, with estimates synthesized using a weighted mean where weights are derived from the Hessian matrix of each bivariate model. Simulation studies demonstrate that the model fits well across different sample sizes. The methodology is illustrated through a real data example, with discussion of forecasting considerations.
