Title: A general theory for detecting changes-in-mean and changes-in-slope
Authors: Chao Zheng - Lancaster University (United Kingdom) [presenting]
Abstract: The aim is to study the finite sample behaviour of an approach to detecting changepoints that is based on maximising a penalised likelihood. These give general results as to when such a procedure can consistently estimate the number of changes and accurately estimate their position. The results we obtained are applied to the problem of detecting changes-in-mean and changes-in-slope. In the latter case we obtain tighter results on the value of penalty that can be used as compared to existing theory. Moreover, the techniques can be easily adapted to other scenarios as long as some basic properties for detecting a single changepoint are satisfied. We demonstrate the usefulness of our approach through numerical experiments on both synthetic data and real data examples.