CMStatistics 2017: Start Registration
View Submission - CFE
Title: Estimating spot variance in possibly explosive AR(1) model: Application to testing bubble with heteroscedastic data Authors:  Dave Harvey - University of Nottingham (United Kingdom)
Steve Leybourne - University of Nottingham (United Kingdom)
Yang Zu - University of Nottingham (United Kingdom) [presenting]
Abstract: Heteroscedasticity, or time-varying volatility/variance, exists widely in economic time series data. Nonparametric estimation is considered for the spot variance function in a varying-coefficient, heteroscedastic AR(1) model, where the AR coefficient of the model dictates a potential change of the model from an unit root regime to an explosive regime. We propose a two-step method to estimate the variance function in the above model. We show the pointwise consistency and the asymptotic normality of the spot variance estimator. We also derive the uniform consistency of the estimator, which is further used in our application in testing explosive bubbles.