Spatial median and self-weighted approach for multivariate heavy-tailed processes
Abstract: A family of multivariate stationary processes is often used to model various data in the fields including econometrics and financial engineering. Some data in econometrics are known to have heavy-tailed property, and for such data, the median regression approach provides a robust estimation procedure in univariate case. This talk proposes a multivariate extension of the median regression approach to multivariate data via so called spatial median regression. We also consider a self-weighting and smoothed empirical likelihood approach to multivariate processes and propose robust inference procedures.

Link aula virtuale Teams