Title: Online instability detection in a nonlinear expectile model: Theoretical and computational aspects Authors:  Matus Maciak - Charles University (Czech Republic) [presenting]
Abstract: An automatic data-driven changepoint detection test is proposed to detect specific instabilities within a nonlinear regression framework. Conditional expectiles, well-known in econometrics for being the only coherent and elicitable risk measure, introduce additional robustness in the underlying model and the proposed statistical test is proved to be consistent while the distribution of the test statistic under the null hypothesis does not depend on the functional form of the underlying model. Resampling techniques are used to obtain the final test decision, and, therefore, relatively easy and straightforward practical application is guaranteed. Important theoretical details are discussed, finite sample empirical properties and real data illustrations are presented.