Dear Statalists,
I have a question about unit root test for very short time series data (only 19 observations, each observation corresponding to one year). I have read from the literature that unit root tests, regardless of ADF or the Phillips-Perron, suffer from size distortion and their power and validity become very weak in very short time series.
So I would like to ask for suggestions about if it is necessary (or alternatively speaking if it makes sense) to do unit root test for such short time series. If not, how should I make a decision on the use of the original (logged) time series (trend-stationary) or differenced series (difference-stationary)?
Many thanks!
I have a question about unit root test for very short time series data (only 19 observations, each observation corresponding to one year). I have read from the literature that unit root tests, regardless of ADF or the Phillips-Perron, suffer from size distortion and their power and validity become very weak in very short time series.
So I would like to ask for suggestions about if it is necessary (or alternatively speaking if it makes sense) to do unit root test for such short time series. If not, how should I make a decision on the use of the original (logged) time series (trend-stationary) or differenced series (difference-stationary)?
Many thanks!
Comment