ARIMA (0,1,0)
Posted: Fri Apr 29, 2011 6:31 am
I have a problem regarding Box jenkins ARIMA method of forecasting
I have exports in US dollars at constant prices . There are 50 0bservations. The series was non stationary as indicated by ADF test. Moreover, ACF correlogram has shown a linear decay to zero whereas PACF shows a signifcant spike at lag 1 and then cut off to zero. After taking first differencing series has became stationary, and ACF, PACF shows no significant spikes. Thus model chosen was ARIMA(0,1,0) a random walk model without drift. However estimating this model yields an output with AR inverted roots greater than 1 and output gives a message that AR(1) is non stationary. why is it happening,although correlogram-Q statistic of residuals test shows no autocorrelation.
I am attaching my excel file for the data series so that one could check my problem
I have exports in US dollars at constant prices . There are 50 0bservations. The series was non stationary as indicated by ADF test. Moreover, ACF correlogram has shown a linear decay to zero whereas PACF shows a signifcant spike at lag 1 and then cut off to zero. After taking first differencing series has became stationary, and ACF, PACF shows no significant spikes. Thus model chosen was ARIMA(0,1,0) a random walk model without drift. However estimating this model yields an output with AR inverted roots greater than 1 and output gives a message that AR(1) is non stationary. why is it happening,although correlogram-Q statistic of residuals test shows no autocorrelation.
I am attaching my excel file for the data series so that one could check my problem