Simulation Analysis of the Size of the BDS Test

Abstract

The BDS test is one of the most important and most commonly used tools for detection of nonlinearity in time series. In the paper, the size of the BDS test is assessed using Monte Carlo simulations. The simulation uses pseudo-random series of different length, generated from seven distributions with different properties. In the research, the approximation of the fi nite sample distribution of the BDS statistic was performed using three methods: the classical one – based on the asymptotic normal distribution and two resampling methods: the bootstrap and the permutation technique.