Modelling Non-Stationary Financial Time Series with Input Warped Student T-Processes

Gheorghe Ruxanda, Sorin Opincariu and Stefan Ionescu
Additional contact information
Gheorghe Ruxanda: Bucharest University of Economic Studies
Sorin Opincariu: Bucharest University of Economic Studies
Stefan Ionescu: Romanian American University.

Journal for Economic Forecasting, 2019, issue 3, 51-61

Abstract: The evolution of financial assets is known to be non-stationary and to present long tails and non-Gaussian. Gaussian processes are a flexible and general Bayesian nonparametric generative model that provide flexible priors on function spaces and interpretable uncertainty quantification. While GP are extremely flexible function approximators, their Gaussian marginal distribution makes them inappropriate to model financial assets returns distributions. We present the Student t-processes that are known to fit heavier tail. We also augment the model with input warping to account with the financial time series non stationarity. We present a case study of fitting the evolution of SP500 index stressing the importance of good uncertainty estimates, especially when the series manifests structural breaks.

Keywords: Bayesian nonparametric; Student processes; Gaussian processes; stylized facts (search for similar items in EconPapers)
JEL-codes: C45 C53 (search for similar items in EconPapers)
Date: 2019
References: Add references at CitEc
Citations: View citations in EconPapers (1)

Downloads: (external link)
http://www.ipe.ro/rjef/rjef3_19/rjef3_2019p51-61.pdf

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:rjr:romjef:v::y:2019:i:3:p:51-61

Access Statistics for this article

Journal for Economic Forecasting is currently edited by Lucian Liviu Albu and Corina Saman

More articles in Journal for Economic Forecasting from Institute for Economic Forecasting Contact information at EDIRC.
Bibliographic data for series maintained by Corina Saman ( this e-mail address is bad, please contact ).