A0480
Title: Real-time changepoint detection in a nonlinear expectile model
Authors: Matus Maciak - Charles University (Czech Republic) [presenting]
Michal Pesta - Charles University (Czech Republic)
Gabriela Ciuperca - University Lyon-I (France)
Abstract: Regime switching within advanced stochastic models attracts a lot of interest over the last years with many different strategies being applied in this direction. We introduce a complex online changepoint detection procedure based on conditional expectiles. Nonlinearity of the underlying model improves the overall flexibility of the overall model, the conditional expectiles---well-known in econometrics for being the only coherent and elicitable risk measure---bring in some additional robustness, and the proposed changepoint detection test is proved to be consistent while the distribution under the null hypothesis depends on neither the functional form of the underlying model nor the unknown parameters which ensure very simple and straightforward applicability for real-life situations. Important theoretical details are summarized and finite sample empirical properties are presented.
