UNIT-ROOTS TEST FOR TIME-SERIES DATA WITH A LINEAR TIME TREND

Time-series models are useful for forecasting, but most standard methods for fitting such models require the series to be stationary. Recently, many tests for detecting unit roots in time-series data (tests for nonstationarity) have been proposed. In this article a test is developed for unit roots in autoregressive, moving-average models containing a linear time trend term. The method of testing is based on the estimation procedure suggested in Fuller (1976), which is basically a nonlinear type of estimation. The test statistic is standard output from most regression programs and has a limit distribution whose percentiles have been tabulated. An illustrative example is given.