Monte Carlo Investigation of Unit Root Tests in Three-Regime SETAR Models

There exists a steadily growing amount of literature concerning statistical inference problems in three-regime SETAR models. Both from theorical and practical points of view, it is crucial to distinguish a linear unit root process from a globally stationary SETAR process. In the last years, several tests have been proposed and their asymptotic null distribution derived. Focusing on supremum and average Wald statistics, we undertake Monte Carlo investigations in order to find critical values of these tests for small samples and to assess their power performance.