Comparison of Unit Root Tests to Examine Stationary Time Series Using Simulation

Abstract

Time series instability is a problem for many studies that may lead to misleading or unrealistic results, so time series stability is a necessary condition in time series analysis. The purpose of this study is to introduce some of the unit root tests (Augmented Dickie Fuller test(ADF) - Philips Peron test (PP) and KPSS test) to detect the stability of the time series and compare them in the examination of the properties of the series and to determine their superiority through the correct classification ratio by simulation using R Package. Results showed that when generating stationary models, the KPSS test is suitable for small and large samples. While the efficiency of the PP and ADF tests is weak for the small samples, the efficiency of the tests increases by increasing the size of the sample. In the generation of the non-stationary models, all KPSS, PP and ADF tests are highly efficient whether the samples are small or large.