The Performance Of Nonlinearity Tests On Asymmetric Nonlinear Time Series

This paper examines nonlinearity tests of asymmetric time series in a controlled Monte Carlo setting with the goal of exploring how well existing nonlinear test statistics performed in a variety of typical time series settings. The data generation processes and sample sizes were allowed to vary in a controlled fashion. The study confirmed that none of the test statistics were dominant relative to the others. Also, dependent upon the data generating process, the test statistics exhibited very different powers. Finally, our research showed that the test performance was heavily dependent upon sample size and the degree of asymmetric mechanism.