On Unit-Root Tests When the Alternative Is a Trend-Break Stationary Process
研究备择假设为趋势断裂平稳过程时单位根检验的统计量，发现若断点形式设定错误会导致严重功效扭曲，建议使用混合模型，并比较了最小t统计量与最大F统计量的表现。
Minimum t statistics to test for a unit-root are available when the form of break under the alternative evolves according to the crash, changing growth, and mixed models. It is shown that serious power distortions occur if the form of break is misspecified, and thus the practitioner should use the mixed model as the appropriate alternative in empirical applications. The mixed model may reveal useful information regarding the location and form of break. The maximum F statistic for the joint null of a unit-root and no breaks is shown to have greater and less erratic power compared to the minimumt statistic. Stronger evidence against the unit-root is found for the Nelson-Plosser series and U.S. Postwar quarterly real gross national product.