arima forecast
Posted: Mon Jan 09, 2012 8:46 pm
hi everyone, will thank heaps a hand on this. i´m trying to forecast using an arima(0,1,1)x(0,1,12). "y" is monthly and in thousands, has unit root and seasonality. i have logged "y" for reducing heterocedasticity, taken 1st diff for trend, and adjusted for seasonality (sma 12). the equation is dlogy = c ma(1) sma(12), where dlogy is seasonal adjusted, all significant except c, high r2 and r2adj, d-w around 2, resids inside the confidence bands. when forcasted (changed sample size, dynamic estimation) the yields are decimals. antilogs are enough for the logs, 1st diff and sma or there r another changes that need to b done to the yiels, besides antilogs, so i get thousands instead of decimals? ta