Time series data is increasingly prevalent, however common time series models can fail to capture the complex dynamics of time series data in practice. In this talk, we focus on a specific model — the ARFIMA(p,d,q) model — and the assumption of stationarity. Assuming that a process is stationary is technically convenient, but may not be appropriate in practice. In this paper, we introduce a likelihood-based approach to estimating the parameters of the popular ARFIMA(p,d,q) model without assuming stationarity. This allows us to implement likelihood-based tests of stationarity and to obtain better estimates of the differencing parameter d.
Maryclare Griffin is an assistant professor of statistics at UMass Amherst. She received a PhD in statistics from the University of Washington in Seattle in 2018 and recently completed a short postdoc at Cornell University. Her research interests include high dimensional regression problems, mixed models, and methods for spatio-temporal data.