Directionality in time series and its applications

Abstract

A suite of seven statistics to detect directionality in time series is presented. Applications from various disciplines including business, environmental science, finance and medicine are considered. Models that allow for directionality are proposed, and methods of fitting these models are investigated. Time series models that incorporate directionality provide more precise prediction limits and more realistic simulations than the models that do not. Potential practical applications include: providing evidence to support physical interpretations; directionality trading rules for investment portfolio; prediction of unstable financial periods; and possible early warning of epileptic seizures.