A0943
Title: Robust LM-type testing in multivariate contaminated time series
Authors: Yukai Yang - Uppsala University (Sweden) [presenting]
Abstract: The challenging problem of performing robust Lagrange Multiplier (LM) tests is addressed in multivariate contaminated time series, where unobservable additive and innovative outliers present significant difficulties. The complexities of detecting and managing outliers within time series models are highlighted. A novel algorithm is introduced, designed to be compatible with any subsample-based LM test, thereby creating a robust version of the LM test. The sufficient conditions required for the algorithm and the modified test are derived to achieve robustness and consistency in the presence of outliers, ensuring reliable performance in large samples. Simulation studies showcase its superior performance and robustness in finite samples. Two real-world applications further demonstrate the practical effectiveness of the algorithm and the resulting robust LM test.
