Efficient algorithms for robust estimation in autoregressive regression models using Student's f distribution

摘要

It is now a common knowledge that most data that occur in practice violate the assumption of normality. The Student's t distribution is not only a handy alternative to such a case but also possesses a structure that makes estimation easy via numerical computation. This paper focuses on maximum-likelihood estimation of the parameters of autoregressive regression model driven by Student's t distribution. EM-type algorithms are utilized to iteratively obtain the maximum-likelihood estimates of the parameters of the model. The method's performance is compared with the method of modified maximum-likelihood estimation in simulations and real data analysis.