A1652
Title: Estimation and testing for multivariate nonlinearity of time series in the presence of additive outliers contamination
Authors: Yukai Yang - Uppsala University (Sweden) [presenting]
Rickard Sandberg - Stockholm School of Economics (Sweden)
Sebastian Ankargren - Uppsala University (Sweden)
Abstract: A robust least-trimmed squares estimator is introduced, designed for multivariate regression analysis of time series data that may be subject to additive outlier contamination, building upon the methodology of a past study. Its associated breakdown point is derived, and its Fisher consistency is established when the error distribution exhibits elliptical symmetry. An expedited algorithm is proposed to improve computational efficiency by applying the C-step procedure in another study. The novel robust estimation method is integrated into the LM-type tests for regime-switching nonlinearity. It is shown that the robust estimator and the robust version of the nonlinearity tests have satisfactory finite sample properties through simulation studies. Two applications illustrate the use of the proposed robust estimator and tests.
