Papers on high-dimensional time series modelling accepted into Journal of the American Statistical Association and Journal of Business & Economic Statistics

Dr Haeran Cho and Dom Owens (COMPASS PhD student), in collaboration with Prof Matteo Barigozzi (Bologna), have had a new paper accepted into Journal of Business & Economic Statistics. Titled ‘FNETS: Factor-adjusted network estimation and forecasting for high-dimensional time series’, the paper proposes a new model for high-dimensional time series data exhibiting dominant dependence, a feature commonly found in many applications including finance and econometrics. The paper further proposes a suite of tools for network analysis and forecasting, which are implemented in the R package ‘fnets’ available on CRAN.

An extension of the model that accommodates multiple structural changes, is proposed in the companion paper written by Dr Haeran Cho in collaboration with Dr Hyeyoung Maeng (Durham), Profs Idris Eckley and Paul Fearnhead (Lancaster). This paper, titled ‘High-dimensional time series segmentation via factor-adjusted vector autoregressive modelling’ and accepted into Journal of the American Statistical Association, develops a methodology for consistent detection and estimation of multiple change points. The utility of the proposed framework is demonstrated in an application to a panel of daily volatilities of the stock prices of US blue chip companies.

The research was funded by Leverhulme Project Grant RPG-2019-390 and the papers (preprints) can be found here and here.