KPSS tests on nonlinear transformations of fractional integration time series: A Monte Carlo studytime series: A Monte Carlo study

• Main Authors: Tsai, Yi-Ting, 蔡佾廷
• Other Authors: Wang, Chien-Ho
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/55953021602805746624

碩士 === 國立臺北大學 === 經濟學系 === 95 === When economists use Dickey-Fuller type unit root tests to detect nonlinear transformations of I(d) series, the test powers of Dickey-Fuller type unit root test are low in general. The main reason to generate low power problem is the characteristic of I(d) series changed after I(d) series have been transformed by some functional form. In this paper, we use KPSS type stationary test to investigate the characteristics of transformed I(d) process. We compare the powers of KPSS statistic and Dickey- Fuller unit root statistic, and try to find the reasons that two statistics have different powers. From Monte Carlo simulation, when , KPSS tests have higher power than traditional Dickey-Fuller type unit root tests. But , KPSS and traditional Dickey-Fuller type unit root tests all have misjudgement for the characteristics of transformed I(d) process. KPSS tests will have the same lower power as Dickey-Fuller unit root tests.