Changepoints are ubiquitous features of climatic time series. Estimation of the number of changepoints and their locations in time series is of great research interests currently. This talk introduces an information based approach to the multiple changepoint estimation (segmentation) problem. As most climate series exhibit serial autocorrelation and monthly, daily, or hourly series may also have periodic mean structures, our methods are specifically tailored to climatic time series in that they allow for periodicities and autocorrelations. The objective function gauging the number of changepoints and their locations is based on a minimum description length (MDL) information criterion, which often provides superior empirical results. A genetic algorithm is then developed to optimize the objective function. The methods are applied in the analysis of a century of monthly temperatures from Tuscaloosa, Alabama.
Keywords: Changepoints; Genetic algorithm; Minimum description length; Periodic autoregressive time series
Biography: Dr. Lu is an Associate Professor of Statistics in the Department of Mathematics and Statistics at Mississippi State University. She received her Ph.D. in Statistics from The University of Georgia in 2004. Dr. Lu's research interests lie broadly in changepoint detection problems, time series analysis, and applications in the environmental sciences.