A new class of statistical methods for analysing long memory time series models with heteroskedasticity

This project will develop new methods of statistical inference for economic and financial time series that exhibit both long memory and heteroskedasticity. Long memory means economic shocks take a long time for their full effect to be felt. Heteroskedasticity means variability changes over time, eg. periods of calm are followed by periods of volatility. While both long memory and heteroskedasticity are commonly found to co-exist in economic time series, there are almost no suitable inference methods. In this research, we will develop a suite of methods for the analysis of long memory that are robust and/or efficient in the presence of heteroskedasticity, and that are applicable in empirically relevant sample sizes.