Hello, I actually have an understanding problem about estimating ARIMA-Models. I use the package forecast and here I simply do best.arima. I can get fitted values by the comand fitted() To estimate the forecasterrors x=c(1:20) y=best.arima(x) fitted(y) forecasterrors(fitted(y), x) My question is, whether this is the right approach or not. Should there be a lag of 1 in y since the forecast is always for the period t+1? I also try to do this procedure for other models (HoltWinters etc) and I want to choose the model with the lowest Mean absolute Error. attached please also find a plot of another arima model and its fitted values. I hope there is an easy answer to my question. Thank you very much in advance and best regards, Markus Schweitzer - http://www.pokertips.tk