21 January 2014 9 1K Report

Traditional unit root tests ignore the presence of breaks in time-series. Few methods are available, like Quandt's statistic, Quandt-Andrews' test, Zivot-Andrews and so on, that assume presence of a "SINGLE' break. If more, say breaks = "TWO", then one could go for, either the AO or IO tests. It is not so uncommon that the breakpoints are more. In that case, what can one do? Is the recent Bai-Perron the only solution ?

More S.J. Balaji's questions See All
Similar questions and discussions